From 9155e70edd8632806c9deb0ad1ac27427538286b Mon Sep 17 00:00:00 2001 From: Peter van der Spek Date: Mon, 10 Aug 2026 11:08:26 +0200 Subject: [PATCH 1/5] feat(melious): add Melious provider with 56 models Melious AI GmbH is a European relay for open-weight models, exposing an OpenAI-compatible endpoint at https://api.melious.ai/v1. - providers/melious: provider.toml, currentColor logo.svg, and 56 model entries (49 chat + 7 embedding), all override-only via base_model. - models/: 16 new lab metadata entries required as base_model targets, under new labs nousresearch, hcompany, baai, intfloat and sentence-transformers. - labs/: descriptions for those five labs, logos for baai, nousresearch and hcompany, plus LAB_NAME_OVERRIDES so they render as "Nous Research", "H Company", "BAAI" and "intfloat" rather than a title-cased directory id. Only models listed on https://melious.ai/hub/models are included. Costs are converted from Melious's EUR list price for the default `balanced` route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). Co-Authored-By: Claude Opus 5 --- labs/baai/lab.toml | 1 + labs/baai/logo.svg | 3 ++ labs/hcompany/lab.toml | 1 + labs/hcompany/logo.svg | 3 ++ labs/intfloat/lab.toml | 1 + labs/nousresearch/lab.toml | 1 + labs/nousresearch/logo.svg | 3 ++ labs/sentence-transformers/lab.toml | 1 + .../alibaba/qwen3-30b-a3b-instruct-2507.toml | 24 +++++++++++++++ models/alibaba/qwen3-embedding-8b.toml | 23 ++++++++++++++ .../alibaba/qwen3-vl-235b-a22b-instruct.toml | 24 +++++++++++++++ models/baai/bge-large-en-v1.5.toml | 23 ++++++++++++++ models/baai/bge-m3.toml | 23 ++++++++++++++ models/baai/bge-multilingual-gemma2.toml | 23 ++++++++++++++ models/google/gemma-3-27b-it.toml | 30 +++++++++++++++++++ models/hcompany/holo2-30b-a3b.toml | 22 ++++++++++++++ .../multilingual-e5-large-instruct.toml | 23 ++++++++++++++ models/intfloat/multilingual-e5-large.toml | 23 ++++++++++++++ models/meta/llama-3.1-405b-instruct.toml | 30 +++++++++++++++++++ models/meta/llama-3.1-8b-instruct.toml | 29 ++++++++++++++++++ models/mistral/voxtral-small-2507.toml | 23 ++++++++++++++ models/nousresearch/hermes-4-405b.toml | 29 ++++++++++++++++++ models/nousresearch/hermes-4-70b.toml | 30 +++++++++++++++++++ ...paraphrase-multilingual-mpnet-base-v2.toml | 23 ++++++++++++++ packages/web/src/render.tsx | 4 +++ providers/melious/logo.svg | 3 ++ .../melious/models/bge-large-en-v1.5.toml | 11 +++++++ providers/melious/models/bge-m3.toml | 8 +++++ .../models/bge-multilingual-gemma2.toml | 8 +++++ .../melious/models/deepseek-r1-0528.toml | 16 ++++++++++ providers/melious/models/deepseek-v3.1.toml | 13 ++++++++ providers/melious/models/deepseek-v3.2.toml | 13 ++++++++ .../models/deepseek-v4-flash-0731.toml | 11 +++++++ .../melious/models/deepseek-v4-flash.toml | 11 +++++++ providers/melious/models/deepseek-v4-pro.toml | 11 +++++++ .../models/devstral-2-123b-instruct-2512.toml | 13 ++++++++ providers/melious/models/gemma-3-27b-it.toml | 11 +++++++ providers/melious/models/gemma-4-26b-a4b.toml | 15 ++++++++++ providers/melious/models/gemma-4-31b.toml | 15 ++++++++++ providers/melious/models/glm-4.7.toml | 10 +++++++ providers/melious/models/glm-5.1.toml | 13 ++++++++ providers/melious/models/glm-5.2.toml | 10 +++++++ providers/melious/models/glm-5.toml | 11 +++++++ providers/melious/models/gpt-oss-120b.toml | 9 ++++++ providers/melious/models/gpt-oss-20b.toml | 9 ++++++ providers/melious/models/hermes-4-405b.toml | 10 +++++++ providers/melious/models/hermes-4-70b.toml | 10 +++++++ providers/melious/models/holo2-30b-a3b.toml | 8 +++++ providers/melious/models/kimi-k2.5.toml | 17 +++++++++++ providers/melious/models/kimi-k2.6.toml | 10 +++++++ providers/melious/models/kimi-k2.7-code.toml | 10 +++++++ providers/melious/models/kimi-k3.toml | 12 ++++++++ .../models/llama-3.1-405b-instruct.toml | 8 +++++ .../melious/models/llama-3.1-8b-instruct.toml | 11 +++++++ .../models/llama-3.3-70b-instruct.toml | 13 ++++++++ providers/melious/models/minimax-m2.5.toml | 13 ++++++++ providers/melious/models/minimax-m2.7.toml | 13 ++++++++ providers/melious/models/minimax-m3.toml | 14 +++++++++ .../models/mistral-medium-3.5-128b.toml | 13 ++++++++ .../mistral-small-3.2-24b-instruct.toml | 13 ++++++++ .../models/mistral-small-4-119b-instruct.toml | 10 +++++++ .../multilingual-e5-large-instruct.toml | 8 +++++ .../melious/models/multilingual-e5-large.toml | 8 +++++ .../models/nemotron-3-nano-30b-a3b.toml | 14 +++++++++ .../models/nemotron-3-super-120b-a12b.toml | 10 +++++++ .../models/paraphrase-multilingual-mpnet.toml | 8 +++++ .../melious/models/pixtral-12b-2409.toml | 9 ++++++ .../models/qwen2.5-vl-72b-instruct.toml | 13 ++++++++ .../models/qwen3-235b-a22b-instruct.toml | 9 ++++++ .../models/qwen3-30b-a3b-instruct.toml | 8 +++++ providers/melious/models/qwen3-32b.toml | 13 ++++++++ .../models/qwen3-coder-30b-a3b-instruct.toml | 9 ++++++ .../melious/models/qwen3-coder-next.toml | 12 ++++++++ .../melious/models/qwen3-embedding-8b.toml | 11 +++++++ .../models/qwen3-next-80b-a3b-thinking.toml | 12 ++++++++ .../models/qwen3-vl-235b-a22b-instruct.toml | 14 +++++++++ .../melious/models/qwen3.5-122b-a10b.toml | 15 ++++++++++ .../melious/models/qwen3.5-397b-a17b.toml | 15 ++++++++++ providers/melious/models/qwen3.5-9b.toml | 13 ++++++++ providers/melious/models/qwen3.6-27b.toml | 12 ++++++++ providers/melious/models/qwen3.6-35b-a3b.toml | 15 ++++++++++ .../models/voxtral-small-24b-2507.toml | 8 +++++ providers/melious/provider.toml | 10 +++++++ 83 files changed, 1072 insertions(+) create mode 100644 labs/baai/lab.toml create mode 100644 labs/baai/logo.svg create mode 100644 labs/hcompany/lab.toml create mode 100644 labs/hcompany/logo.svg create mode 100644 labs/intfloat/lab.toml create mode 100644 labs/nousresearch/lab.toml create mode 100644 labs/nousresearch/logo.svg create mode 100644 labs/sentence-transformers/lab.toml create mode 100644 models/alibaba/qwen3-30b-a3b-instruct-2507.toml create mode 100644 models/alibaba/qwen3-embedding-8b.toml create mode 100644 models/alibaba/qwen3-vl-235b-a22b-instruct.toml create mode 100644 models/baai/bge-large-en-v1.5.toml create mode 100644 models/baai/bge-m3.toml create mode 100644 models/baai/bge-multilingual-gemma2.toml create mode 100644 models/google/gemma-3-27b-it.toml create mode 100644 models/hcompany/holo2-30b-a3b.toml create mode 100644 models/intfloat/multilingual-e5-large-instruct.toml create mode 100644 models/intfloat/multilingual-e5-large.toml create mode 100644 models/meta/llama-3.1-405b-instruct.toml create mode 100644 models/meta/llama-3.1-8b-instruct.toml create mode 100644 models/mistral/voxtral-small-2507.toml create mode 100644 models/nousresearch/hermes-4-405b.toml create mode 100644 models/nousresearch/hermes-4-70b.toml create mode 100644 models/sentence-transformers/paraphrase-multilingual-mpnet-base-v2.toml create mode 100644 providers/melious/logo.svg create mode 100644 providers/melious/models/bge-large-en-v1.5.toml create mode 100644 providers/melious/models/bge-m3.toml create mode 100644 providers/melious/models/bge-multilingual-gemma2.toml create mode 100644 providers/melious/models/deepseek-r1-0528.toml create mode 100644 providers/melious/models/deepseek-v3.1.toml create mode 100644 providers/melious/models/deepseek-v3.2.toml create mode 100644 providers/melious/models/deepseek-v4-flash-0731.toml create mode 100644 providers/melious/models/deepseek-v4-flash.toml create mode 100644 providers/melious/models/deepseek-v4-pro.toml create mode 100644 providers/melious/models/devstral-2-123b-instruct-2512.toml create mode 100644 providers/melious/models/gemma-3-27b-it.toml create mode 100644 providers/melious/models/gemma-4-26b-a4b.toml create mode 100644 providers/melious/models/gemma-4-31b.toml create mode 100644 providers/melious/models/glm-4.7.toml create mode 100644 providers/melious/models/glm-5.1.toml create mode 100644 providers/melious/models/glm-5.2.toml create mode 100644 providers/melious/models/glm-5.toml create mode 100644 providers/melious/models/gpt-oss-120b.toml create mode 100644 providers/melious/models/gpt-oss-20b.toml create mode 100644 providers/melious/models/hermes-4-405b.toml create mode 100644 providers/melious/models/hermes-4-70b.toml create mode 100644 providers/melious/models/holo2-30b-a3b.toml create mode 100644 providers/melious/models/kimi-k2.5.toml create mode 100644 providers/melious/models/kimi-k2.6.toml create mode 100644 providers/melious/models/kimi-k2.7-code.toml create mode 100644 providers/melious/models/kimi-k3.toml create mode 100644 providers/melious/models/llama-3.1-405b-instruct.toml create mode 100644 providers/melious/models/llama-3.1-8b-instruct.toml create mode 100644 providers/melious/models/llama-3.3-70b-instruct.toml create mode 100644 providers/melious/models/minimax-m2.5.toml create mode 100644 providers/melious/models/minimax-m2.7.toml create mode 100644 providers/melious/models/minimax-m3.toml create mode 100644 providers/melious/models/mistral-medium-3.5-128b.toml create mode 100644 providers/melious/models/mistral-small-3.2-24b-instruct.toml create mode 100644 providers/melious/models/mistral-small-4-119b-instruct.toml create mode 100644 providers/melious/models/multilingual-e5-large-instruct.toml create mode 100644 providers/melious/models/multilingual-e5-large.toml create mode 100644 providers/melious/models/nemotron-3-nano-30b-a3b.toml create mode 100644 providers/melious/models/nemotron-3-super-120b-a12b.toml create mode 100644 providers/melious/models/paraphrase-multilingual-mpnet.toml create mode 100644 providers/melious/models/pixtral-12b-2409.toml create mode 100644 providers/melious/models/qwen2.5-vl-72b-instruct.toml create mode 100644 providers/melious/models/qwen3-235b-a22b-instruct.toml create mode 100644 providers/melious/models/qwen3-30b-a3b-instruct.toml create mode 100644 providers/melious/models/qwen3-32b.toml create mode 100644 providers/melious/models/qwen3-coder-30b-a3b-instruct.toml create mode 100644 providers/melious/models/qwen3-coder-next.toml create mode 100644 providers/melious/models/qwen3-embedding-8b.toml create mode 100644 providers/melious/models/qwen3-next-80b-a3b-thinking.toml create mode 100644 providers/melious/models/qwen3-vl-235b-a22b-instruct.toml create mode 100644 providers/melious/models/qwen3.5-122b-a10b.toml create mode 100644 providers/melious/models/qwen3.5-397b-a17b.toml create mode 100644 providers/melious/models/qwen3.5-9b.toml create mode 100644 providers/melious/models/qwen3.6-27b.toml create mode 100644 providers/melious/models/qwen3.6-35b-a3b.toml create mode 100644 providers/melious/models/voxtral-small-24b-2507.toml create mode 100644 providers/melious/provider.toml diff --git a/labs/baai/lab.toml b/labs/baai/lab.toml new file mode 100644 index 00000000000..25deed35af8 --- /dev/null +++ b/labs/baai/lab.toml @@ -0,0 +1 @@ +description = "The Beijing Academy of Artificial Intelligence publishes the BGE family, a widely used set of open embedding and reranking models for retrieval." diff --git a/labs/baai/logo.svg b/labs/baai/logo.svg new file mode 100644 index 00000000000..60a5e071fa6 --- /dev/null +++ b/labs/baai/logo.svg @@ -0,0 +1,3 @@ + + + diff --git a/labs/hcompany/lab.toml b/labs/hcompany/lab.toml new file mode 100644 index 00000000000..7d770a9bd2e --- /dev/null +++ b/labs/hcompany/lab.toml @@ -0,0 +1 @@ +description = "H Company builds Holo models for GUI grounding, letting computer-use agents locate and act on elements in web, desktop, and mobile interfaces." diff --git a/labs/hcompany/logo.svg b/labs/hcompany/logo.svg new file mode 100644 index 00000000000..e49d7f70705 --- /dev/null +++ b/labs/hcompany/logo.svg @@ -0,0 +1,3 @@ + + + diff --git a/labs/intfloat/lab.toml b/labs/intfloat/lab.toml new file mode 100644 index 00000000000..2150a25aa82 --- /dev/null +++ b/labs/intfloat/lab.toml @@ -0,0 +1 @@ +description = "intfloat publishes the E5 text-embedding family, contrastively trained retrieval models covering English and multilingual corpora." diff --git a/labs/nousresearch/lab.toml b/labs/nousresearch/lab.toml new file mode 100644 index 00000000000..f32373c0344 --- /dev/null +++ b/labs/nousresearch/lab.toml @@ -0,0 +1 @@ +description = "Nous Research fine-tunes open base models into the Hermes line, favouring steerable, hybrid-reasoning assistants with neutral alignment." diff --git a/labs/nousresearch/logo.svg b/labs/nousresearch/logo.svg new file mode 100644 index 00000000000..979c67878a4 --- /dev/null +++ b/labs/nousresearch/logo.svg @@ -0,0 +1,3 @@ + + + diff --git a/labs/sentence-transformers/lab.toml b/labs/sentence-transformers/lab.toml new file mode 100644 index 00000000000..96207b4a9d0 --- /dev/null +++ b/labs/sentence-transformers/lab.toml @@ -0,0 +1 @@ +description = "The sentence-transformers project maintains compact bi-encoder embedding models for sentence similarity, clustering, and paraphrase mining." diff --git a/models/alibaba/qwen3-30b-a3b-instruct-2507.toml b/models/alibaba/qwen3-30b-a3b-instruct-2507.toml new file mode 100644 index 00000000000..3f1c8dada1d --- /dev/null +++ b/models/alibaba/qwen3-30b-a3b-instruct-2507.toml @@ -0,0 +1,24 @@ +name = "Qwen3 30B-A3B Instruct 2507" +description = "Compact open Qwen3 MoE instruct model for fast multilingual chat, coding, and tool use" +family = "qwen" +release_date = "2025-07-29" +last_updated = "2025-07-29" +attachment = false +reasoning = false +temperature = true +tool_call = true +structured_output = true +open_weights = true +license = "Apache-2.0" + +[limit] +context = 262_144 +output = 16_384 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507" diff --git a/models/alibaba/qwen3-embedding-8b.toml b/models/alibaba/qwen3-embedding-8b.toml new file mode 100644 index 00000000000..e4e5bb66629 --- /dev/null +++ b/models/alibaba/qwen3-embedding-8b.toml @@ -0,0 +1,23 @@ +name = "Qwen3 Embedding 8B" +description = "Largest Qwen3 embedding model for multilingual retrieval, ranking, and code search" +family = "qwen" +release_date = "2025-06-03" +last_updated = "2025-06-03" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "Apache-2.0" + +[limit] +context = 32_768 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/Qwen/Qwen3-Embedding-8B" diff --git a/models/alibaba/qwen3-vl-235b-a22b-instruct.toml b/models/alibaba/qwen3-vl-235b-a22b-instruct.toml new file mode 100644 index 00000000000..2f7905c522a --- /dev/null +++ b/models/alibaba/qwen3-vl-235b-a22b-instruct.toml @@ -0,0 +1,24 @@ +name = "Qwen3 VL 235B-A22B Instruct" +description = "Large open Qwen3 vision-language MoE model for document, chart, and screenshot understanding" +family = "qwen" +release_date = "2025-09-23" +last_updated = "2025-09-23" +attachment = true +reasoning = false +temperature = true +tool_call = true +structured_output = true +open_weights = true +license = "Apache-2.0" + +[limit] +context = 262_144 +output = 32_768 + +[modalities] +input = ["text", "image", "video"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Instruct" diff --git a/models/baai/bge-large-en-v1.5.toml b/models/baai/bge-large-en-v1.5.toml new file mode 100644 index 00000000000..08f3be49f70 --- /dev/null +++ b/models/baai/bge-large-en-v1.5.toml @@ -0,0 +1,23 @@ +name = "BGE Large EN v1.5" +description = "English text embedding model for semantic search, retrieval, and reranking pipelines" +family = "bge" +release_date = "2023-09-12" +last_updated = "2023-09-12" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "MIT" + +[limit] +context = 512 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/BAAI/bge-large-en-v1.5" diff --git a/models/baai/bge-m3.toml b/models/baai/bge-m3.toml new file mode 100644 index 00000000000..b17042c4fe5 --- /dev/null +++ b/models/baai/bge-m3.toml @@ -0,0 +1,23 @@ +name = "BGE M3" +description = "Multilingual multi-granularity embedding model for dense, sparse, and multi-vector retrieval" +family = "bge" +release_date = "2024-01-27" +last_updated = "2024-01-27" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "MIT" + +[limit] +context = 8_192 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/BAAI/bge-m3" diff --git a/models/baai/bge-multilingual-gemma2.toml b/models/baai/bge-multilingual-gemma2.toml new file mode 100644 index 00000000000..c263c90e792 --- /dev/null +++ b/models/baai/bge-multilingual-gemma2.toml @@ -0,0 +1,23 @@ +name = "BGE Multilingual Gemma2" +description = "Gemma 2 based multilingual embedding model for cross-language search and retrieval" +family = "bge" +release_date = "2024-07-29" +last_updated = "2024-07-29" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "Gemma" + +[limit] +context = 8_192 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/BAAI/bge-multilingual-gemma2" diff --git a/models/google/gemma-3-27b-it.toml b/models/google/gemma-3-27b-it.toml new file mode 100644 index 00000000000..c34962dc7e9 --- /dev/null +++ b/models/google/gemma-3-27b-it.toml @@ -0,0 +1,30 @@ +name = "Gemma 3 27B IT" +description = "Open Gemma 3 instruction model with vision input and a long multilingual context" +family = "gemma" +release_date = "2025-03-12" +last_updated = "2025-03-12" +attachment = true +reasoning = false +temperature = true +tool_call = true +structured_output = true +knowledge = "2024-08" +open_weights = true +license = "Gemma" + +[limit] +context = 131_072 +output = 8_192 + +[modalities] +input = ["text", "image"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/google/gemma-3-27b-it" + +[[links]] +label = "Model card" +url = "https://ai.google.dev/gemma/docs/core/model_card_3" +type = "model_card" diff --git a/models/hcompany/holo2-30b-a3b.toml b/models/hcompany/holo2-30b-a3b.toml new file mode 100644 index 00000000000..86d356526bc --- /dev/null +++ b/models/hcompany/holo2-30b-a3b.toml @@ -0,0 +1,22 @@ +name = "Holo2 30B A3B" +description = "Open GUI-grounding model that reads web, desktop, and mobile interfaces for computer-use agents" +release_date = "2025-11-10" +last_updated = "2025-11-10" +attachment = true +reasoning = false +temperature = true +tool_call = true +open_weights = true +license = "Apache-2.0" + +[limit] +context = 131_072 +output = 16_384 + +[modalities] +input = ["text", "image"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/Hcompany/Holo2-30B-A3B" diff --git a/models/intfloat/multilingual-e5-large-instruct.toml b/models/intfloat/multilingual-e5-large-instruct.toml new file mode 100644 index 00000000000..3acda9639c7 --- /dev/null +++ b/models/intfloat/multilingual-e5-large-instruct.toml @@ -0,0 +1,23 @@ +name = "Multilingual E5 Large Instruct" +description = "Instruction-tuned multilingual E5 embedding model for task-conditioned retrieval" +family = "text-embedding" +release_date = "2024-02-08" +last_updated = "2024-02-08" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "MIT" + +[limit] +context = 512 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/intfloat/multilingual-e5-large-instruct" diff --git a/models/intfloat/multilingual-e5-large.toml b/models/intfloat/multilingual-e5-large.toml new file mode 100644 index 00000000000..bb3ea0a6f03 --- /dev/null +++ b/models/intfloat/multilingual-e5-large.toml @@ -0,0 +1,23 @@ +name = "Multilingual E5 Large" +description = "Multilingual E5 embedding model for semantic search and retrieval across 100 languages" +family = "text-embedding" +release_date = "2023-06-30" +last_updated = "2023-06-30" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "MIT" + +[limit] +context = 512 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/intfloat/multilingual-e5-large" diff --git a/models/meta/llama-3.1-405b-instruct.toml b/models/meta/llama-3.1-405b-instruct.toml new file mode 100644 index 00000000000..90eccf57c64 --- /dev/null +++ b/models/meta/llama-3.1-405b-instruct.toml @@ -0,0 +1,30 @@ +name = "Llama 3.1 405B Instruct" +description = "Largest open Llama 3.1 instruct model for multilingual chat, reasoning, and coding" +family = "llama" +release_date = "2024-07-23" +last_updated = "2024-07-23" +attachment = false +reasoning = false +temperature = true +tool_call = true +structured_output = true +knowledge = "2023-12" +open_weights = true +license = "Llama 3.1 Community License" + +[limit] +context = 128_000 +output = 4_096 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct" + +[[links]] +label = "Model card" +url = "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md" +type = "model_card" diff --git a/models/meta/llama-3.1-8b-instruct.toml b/models/meta/llama-3.1-8b-instruct.toml new file mode 100644 index 00000000000..72cd07941c1 --- /dev/null +++ b/models/meta/llama-3.1-8b-instruct.toml @@ -0,0 +1,29 @@ +name = "Llama 3.1 8B Instruct" +description = "Small open Llama 3.1 instruct model for cheap multilingual chat and self-hosting" +family = "llama" +release_date = "2024-07-23" +last_updated = "2024-07-23" +attachment = false +reasoning = false +temperature = true +tool_call = true +knowledge = "2023-12" +open_weights = true +license = "Llama 3.1 Community License" + +[limit] +context = 131_072 +output = 4_096 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct" + +[[links]] +label = "Model card" +url = "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md" +type = "model_card" diff --git a/models/mistral/voxtral-small-2507.toml b/models/mistral/voxtral-small-2507.toml new file mode 100644 index 00000000000..ec6b9c98a0b --- /dev/null +++ b/models/mistral/voxtral-small-2507.toml @@ -0,0 +1,23 @@ +name = "Voxtral Small 24B 2507" +description = "Open Mistral speech-understanding model for transcription, audio Q&A, and summarisation" +family = "voxtral" +release_date = "2025-07-15" +last_updated = "2025-07-15" +attachment = true +reasoning = false +temperature = true +tool_call = false +open_weights = true +license = "Apache-2.0" + +[limit] +context = 32_000 +output = 4_096 + +[modalities] +input = ["text", "audio"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/mistralai/Voxtral-Small-24B-2507" diff --git a/models/nousresearch/hermes-4-405b.toml b/models/nousresearch/hermes-4-405b.toml new file mode 100644 index 00000000000..f127c2c7bdc --- /dev/null +++ b/models/nousresearch/hermes-4-405b.toml @@ -0,0 +1,29 @@ +name = "Hermes 4 405B" +description = "Hybrid-reasoning Hermes 4 built on Llama 3.1 405B for logic, math, and steerable chat" +family = "hermes" +release_date = "2025-08-26" +last_updated = "2025-08-26" +attachment = false +reasoning = true +temperature = true +tool_call = true +knowledge = "2023-12" +open_weights = true +license = "Llama 3.1 Community License" + +[limit] +context = 128_000 +output = 16_384 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/NousResearch/Hermes-4-405B" + +[[links]] +label = "Announcement" +url = "https://hermes4.nousresearch.com" +type = "announcement" diff --git a/models/nousresearch/hermes-4-70b.toml b/models/nousresearch/hermes-4-70b.toml new file mode 100644 index 00000000000..4fa2ec02dd6 --- /dev/null +++ b/models/nousresearch/hermes-4-70b.toml @@ -0,0 +1,30 @@ +name = "Hermes 4 70B" +description = "Hybrid-reasoning Hermes 4 built on Llama 3.1 70B for deliberate analysis and tool use" +family = "hermes" +release_date = "2025-08-26" +last_updated = "2025-08-26" +attachment = false +reasoning = true +temperature = true +tool_call = true +structured_output = true +knowledge = "2023-12" +open_weights = true +license = "Llama 3.1 Community License" + +[limit] +context = 128_000 +output = 16_384 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/NousResearch/Hermes-4-70B" + +[[links]] +label = "Announcement" +url = "https://hermes4.nousresearch.com" +type = "announcement" diff --git a/models/sentence-transformers/paraphrase-multilingual-mpnet-base-v2.toml b/models/sentence-transformers/paraphrase-multilingual-mpnet-base-v2.toml new file mode 100644 index 00000000000..d855ffd6d52 --- /dev/null +++ b/models/sentence-transformers/paraphrase-multilingual-mpnet-base-v2.toml @@ -0,0 +1,23 @@ +name = "Paraphrase Multilingual MPNet Base v2" +description = "Compact multilingual sentence embedding model for paraphrase matching and clustering" +family = "text-embedding" +release_date = "2021-06-02" +last_updated = "2021-06-02" +attachment = false +reasoning = false +temperature = false +tool_call = false +open_weights = true +license = "Apache-2.0" + +[limit] +context = 128 +output = 1 + +[modalities] +input = ["text"] +output = ["text"] + +[[weights]] +label = "Hugging Face" +url = "https://huggingface.co/sentence-transformers/paraphrase-multilingual-mpnet-base-v2" diff --git a/packages/web/src/render.tsx b/packages/web/src/render.tsx index 93c73c6e66e..331cfe8701d 100644 --- a/packages/web/src/render.tsx +++ b/packages/web/src/render.tsx @@ -99,9 +99,13 @@ interface SearchIndexItem { const LAB_NAME_OVERRIDES: Record = { alibaba: "Alibaba", + baai: "BAAI", + hcompany: "H Company", + intfloat: "intfloat", meta: "Meta", minimax: "MiniMax", moonshotai: "Moonshot AI", + nousresearch: "Nous Research", openai: "OpenAI", perplexity: "Perplexity", stepfun: "StepFun", diff --git a/providers/melious/logo.svg b/providers/melious/logo.svg new file mode 100644 index 00000000000..a2ff326807a --- /dev/null +++ b/providers/melious/logo.svg @@ -0,0 +1,3 @@ + + + diff --git a/providers/melious/models/bge-large-en-v1.5.toml b/providers/melious/models/bge-large-en-v1.5.toml new file mode 100644 index 00000000000..103c5b19f12 --- /dev/null +++ b/providers/melious/models/bge-large-en-v1.5.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/bge-large-en-v1.5 and https://melious.ai/pricing +base_model = "baai/bge-large-en-v1.5" + +[cost] +input = 0.0231 +output = 0 + +[limit] +context = 8192 diff --git a/providers/melious/models/bge-m3.toml b/providers/melious/models/bge-m3.toml new file mode 100644 index 00000000000..3a2104840fe --- /dev/null +++ b/providers/melious/models/bge-m3.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/bge-m3 and https://melious.ai/pricing +base_model = "baai/bge-m3" + +[cost] +input = 0.0115 +output = 0 diff --git a/providers/melious/models/bge-multilingual-gemma2.toml b/providers/melious/models/bge-multilingual-gemma2.toml new file mode 100644 index 00000000000..fb07c432173 --- /dev/null +++ b/providers/melious/models/bge-multilingual-gemma2.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/bge-multilingual-gemma2 and https://melious.ai/pricing +base_model = "baai/bge-multilingual-gemma2" + +[cost] +input = 0.0115 +output = 0 diff --git a/providers/melious/models/deepseek-r1-0528.toml b/providers/melious/models/deepseek-r1-0528.toml new file mode 100644 index 00000000000..8d34fd98f34 --- /dev/null +++ b/providers/melious/models/deepseek-r1-0528.toml @@ -0,0 +1,16 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/deepseek-r1-0528 and https://melious.ai/pricing +# R1 always reasons and exposes no effort control; served as the 0528 checkpoint. +base_model = "deepseek/deepseek-r1" +name = "DeepSeek R1 0528" +structured_output = true +reasoning_options = [] + +[cost] +input = 0.7498 +output = 2.9991 +cache_read = 0.1846 + +[limit] +context = 164_000 diff --git a/providers/melious/models/deepseek-v3.1.toml b/providers/melious/models/deepseek-v3.1.toml new file mode 100644 index 00000000000..3384b5e1bdb --- /dev/null +++ b/providers/melious/models/deepseek-v3.1.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/deepseek-v3.1 and https://melious.ai/pricing +base_model = "deepseek/deepseek-v3.1" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.2307 +output = 0.9228 + +[limit] +context = 164_000 diff --git a/providers/melious/models/deepseek-v3.2.toml b/providers/melious/models/deepseek-v3.2.toml new file mode 100644 index 00000000000..260d66b711e --- /dev/null +++ b/providers/melious/models/deepseek-v3.2.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/deepseek-v3.2 and https://melious.ai/pricing +base_model = "deepseek/deepseek-v3.2" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.346 +output = 0.5767 +cache_read = 0.0923 + +[limit] +context = 164_000 diff --git a/providers/melious/models/deepseek-v4-flash-0731.toml b/providers/melious/models/deepseek-v4-flash-0731.toml new file mode 100644 index 00000000000..d1dd5455ea4 --- /dev/null +++ b/providers/melious/models/deepseek-v4-flash-0731.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/deepseek-v4-flash-0731 and https://melious.ai/pricing +# Effort: DeepSeek V4 collapses low/medium onto high, so only high and max differ. +base_model = "deepseek/deepseek-v4-flash-0731" +reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }] + +[cost] +input = 0.2884 +output = 0.346 +cache_read = 0.0692 diff --git a/providers/melious/models/deepseek-v4-flash.toml b/providers/melious/models/deepseek-v4-flash.toml new file mode 100644 index 00000000000..5a8e66bd40e --- /dev/null +++ b/providers/melious/models/deepseek-v4-flash.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/deepseek-v4-flash and https://melious.ai/pricing +# Effort: DeepSeek V4 collapses low/medium onto high, so only high and max differ. +base_model = "deepseek/deepseek-v4-flash" +reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }] + +[cost] +input = 0.173 +output = 0.346 +cache_read = 0.0461 diff --git a/providers/melious/models/deepseek-v4-pro.toml b/providers/melious/models/deepseek-v4-pro.toml new file mode 100644 index 00000000000..0b47544b65a --- /dev/null +++ b/providers/melious/models/deepseek-v4-pro.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/deepseek-v4-pro and https://melious.ai/pricing +# Effort: DeepSeek V4 collapses low/medium onto high, so only high and max differ. +base_model = "deepseek/deepseek-v4-pro" +reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }] + +[cost] +input = 1.8456 +output = 3.6912 +cache_read = 0.4729 diff --git a/providers/melious/models/devstral-2-123b-instruct-2512.toml b/providers/melious/models/devstral-2-123b-instruct-2512.toml new file mode 100644 index 00000000000..9b6e41493e2 --- /dev/null +++ b/providers/melious/models/devstral-2-123b-instruct-2512.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/devstral-2-123b-instruct-2512 and https://melious.ai/pricing +base_model = "mistral/devstral-2512" +structured_output = true + +[cost] +input = 0.4614 +output = 2.307 + +[limit] +context = 200_000 +output = 200_000 diff --git a/providers/melious/models/gemma-3-27b-it.toml b/providers/melious/models/gemma-3-27b-it.toml new file mode 100644 index 00000000000..6441a118440 --- /dev/null +++ b/providers/melious/models/gemma-3-27b-it.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/gemma-3-27b-it and https://melious.ai/pricing +base_model = "google/gemma-3-27b-it" + +[cost] +input = 0.2884 +output = 0.5767 + +[limit] +context = 40_000 diff --git a/providers/melious/models/gemma-4-26b-a4b.toml b/providers/melious/models/gemma-4-26b-a4b.toml new file mode 100644 index 00000000000..dafd7234515 --- /dev/null +++ b/providers/melious/models/gemma-4-26b-a4b.toml @@ -0,0 +1,15 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/gemma-4-26b-a4b and https://melious.ai/pricing +base_model = "google/gemma-4-26b-a4b-it" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.1154 +output = 0.5767 + +[limit] +context = 256_000 + +[modalities] +input = ["text", "image", "video"] diff --git a/providers/melious/models/gemma-4-31b.toml b/providers/melious/models/gemma-4-31b.toml new file mode 100644 index 00000000000..dd115c5f951 --- /dev/null +++ b/providers/melious/models/gemma-4-31b.toml @@ -0,0 +1,15 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/gemma-4-31b and https://melious.ai/pricing +base_model = "google/gemma-4-31b-it" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.1154 +output = 0.346 + +[limit] +context = 256_000 + +[modalities] +input = ["text", "image", "video"] diff --git a/providers/melious/models/glm-4.7.toml b/providers/melious/models/glm-4.7.toml new file mode 100644 index 00000000000..73e61f23916 --- /dev/null +++ b/providers/melious/models/glm-4.7.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/glm-4.7 and https://melious.ai/pricing +base_model = "zhipuai/glm-4.7" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.8074 +output = 3.2298 diff --git a/providers/melious/models/glm-5.1.toml b/providers/melious/models/glm-5.1.toml new file mode 100644 index 00000000000..16feb6fa624 --- /dev/null +++ b/providers/melious/models/glm-5.1.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/glm-5.1 and https://melious.ai/pricing +base_model = "zhipuai/glm-5.1" +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 1.4995 +output = 4.6717 +cache_read = 0.4037 + +[limit] +context = 203_000 diff --git a/providers/melious/models/glm-5.2.toml b/providers/melious/models/glm-5.2.toml new file mode 100644 index 00000000000..ea757e61d8e --- /dev/null +++ b/providers/melious/models/glm-5.2.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/glm-5.2 and https://melious.ai/pricing +base_model = "zhipuai/glm-5.2" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 1.1535 +output = 4.614 +cache_read = 0.2884 diff --git a/providers/melious/models/glm-5.toml b/providers/melious/models/glm-5.toml new file mode 100644 index 00000000000..192de4b0168 --- /dev/null +++ b/providers/melious/models/glm-5.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/glm-5 and https://melious.ai/pricing +base_model = "zhipuai/glm-5" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 1.0958 +output = 3.4028 +cache_read = 0.2884 diff --git a/providers/melious/models/gpt-oss-120b.toml b/providers/melious/models/gpt-oss-120b.toml new file mode 100644 index 00000000000..dea91a092c6 --- /dev/null +++ b/providers/melious/models/gpt-oss-120b.toml @@ -0,0 +1,9 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/gpt-oss-120b and https://melious.ai/pricing +base_model = "openai/gpt-oss-120b" +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] + +[cost] +input = 0.2538 +output = 0.7613 diff --git a/providers/melious/models/gpt-oss-20b.toml b/providers/melious/models/gpt-oss-20b.toml new file mode 100644 index 00000000000..a61f7655436 --- /dev/null +++ b/providers/melious/models/gpt-oss-20b.toml @@ -0,0 +1,9 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/gpt-oss-20b and https://melious.ai/pricing +base_model = "openai/gpt-oss-20b" +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] + +[cost] +input = 0.0461 +output = 0.1846 diff --git a/providers/melious/models/hermes-4-405b.toml b/providers/melious/models/hermes-4-405b.toml new file mode 100644 index 00000000000..7e6b3dac15e --- /dev/null +++ b/providers/melious/models/hermes-4-405b.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/hermes-4-405b and https://melious.ai/pricing +# Hermes 4 hybrid reasoning is steered by the system prompt, not a request field. +base_model = "nousresearch/hermes-4-405b" +reasoning_options = [] + +[cost] +input = 1.0958 +output = 3.2875 diff --git a/providers/melious/models/hermes-4-70b.toml b/providers/melious/models/hermes-4-70b.toml new file mode 100644 index 00000000000..f30144d603e --- /dev/null +++ b/providers/melious/models/hermes-4-70b.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/hermes-4-70b and https://melious.ai/pricing +# Hermes 4 hybrid reasoning is steered by the system prompt, not a request field. +base_model = "nousresearch/hermes-4-70b" +reasoning_options = [] + +[cost] +input = 0.1384 +output = 0.4614 diff --git a/providers/melious/models/holo2-30b-a3b.toml b/providers/melious/models/holo2-30b-a3b.toml new file mode 100644 index 00000000000..f9a225e796b --- /dev/null +++ b/providers/melious/models/holo2-30b-a3b.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/holo2-30b-a3b and https://melious.ai/pricing +base_model = "hcompany/holo2-30b-a3b" + +[cost] +input = 0.346 +output = 0.8074 diff --git a/providers/melious/models/kimi-k2.5.toml b/providers/melious/models/kimi-k2.5.toml new file mode 100644 index 00000000000..41f9aae016f --- /dev/null +++ b/providers/melious/models/kimi-k2.5.toml @@ -0,0 +1,17 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/kimi-k2.5 and https://melious.ai/pricing +base_model = "moonshotai/kimi-k2.5" +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.5767 +output = 2.9414 +cache_read = 0.1384 + +[limit] +context = 256_000 +output = 256_000 + +[modalities] +input = ["text", "image"] diff --git a/providers/melious/models/kimi-k2.6.toml b/providers/melious/models/kimi-k2.6.toml new file mode 100644 index 00000000000..d5190059a7b --- /dev/null +++ b/providers/melious/models/kimi-k2.6.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/kimi-k2.6 and https://melious.ai/pricing +base_model = "moonshotai/kimi-k2.6" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.8074 +output = 4.0373 +cache_read = 0.2884 diff --git a/providers/melious/models/kimi-k2.7-code.toml b/providers/melious/models/kimi-k2.7-code.toml new file mode 100644 index 00000000000..827555cacf2 --- /dev/null +++ b/providers/melious/models/kimi-k2.7-code.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/kimi-k2.7-code and https://melious.ai/pricing +base_model = "moonshotai/kimi-k2.7-code" +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] + +[cost] +input = 0.8074 +output = 4.0373 +cache_read = 0.2307 diff --git a/providers/melious/models/kimi-k3.toml b/providers/melious/models/kimi-k3.toml new file mode 100644 index 00000000000..8bf9fd588dd --- /dev/null +++ b/providers/melious/models/kimi-k3.toml @@ -0,0 +1,12 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/kimi-k3 and https://melious.ai/pricing +base_model = "moonshotai/kimi-k3" +reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }] + +[cost] +input = 3.4605 +output = 17.3025 + +[limit] +context = 1_000_000 diff --git a/providers/melious/models/llama-3.1-405b-instruct.toml b/providers/melious/models/llama-3.1-405b-instruct.toml new file mode 100644 index 00000000000..7c0b3655bdd --- /dev/null +++ b/providers/melious/models/llama-3.1-405b-instruct.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/llama-3.1-405b-instruct and https://melious.ai/pricing +base_model = "meta/llama-3.1-405b-instruct" + +[cost] +input = 2.0186 +output = 2.0186 diff --git a/providers/melious/models/llama-3.1-8b-instruct.toml b/providers/melious/models/llama-3.1-8b-instruct.toml new file mode 100644 index 00000000000..de5364e122d --- /dev/null +++ b/providers/melious/models/llama-3.1-8b-instruct.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/llama-3.1-8b-instruct and https://melious.ai/pricing +base_model = "meta/llama-3.1-8b-instruct" + +[cost] +input = 0.173 +output = 0.173 + +[limit] +context = 128_000 diff --git a/providers/melious/models/llama-3.3-70b-instruct.toml b/providers/melious/models/llama-3.3-70b-instruct.toml new file mode 100644 index 00000000000..e753a1f90eb --- /dev/null +++ b/providers/melious/models/llama-3.3-70b-instruct.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/llama-3.3-70b-instruct and https://melious.ai/pricing +base_model = "meta/llama-3.3-70b-instruct" +attachment = false +structured_output = true + +[cost] +input = 0.7498 +output = 0.7498 + +[limit] +context = 131_072 diff --git a/providers/melious/models/minimax-m2.5.toml b/providers/melious/models/minimax-m2.5.toml new file mode 100644 index 00000000000..915a15b3cf8 --- /dev/null +++ b/providers/melious/models/minimax-m2.5.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/minimax-m2.5 and https://melious.ai/pricing +base_model = "minimax/MiniMax-M2.5" +structured_output = true +reasoning_options = [] + +[cost] +input = 0.2999 +output = 1.1996 + +[limit] +context = 197_000 diff --git a/providers/melious/models/minimax-m2.7.toml b/providers/melious/models/minimax-m2.7.toml new file mode 100644 index 00000000000..6fc55563ed8 --- /dev/null +++ b/providers/melious/models/minimax-m2.7.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/minimax-m2.7 and https://melious.ai/pricing +base_model = "minimax/MiniMax-M2.7" +structured_output = true +reasoning_options = [] + +[cost] +input = 0.6921 +output = 2.7684 + +[limit] +context = 192_000 diff --git a/providers/melious/models/minimax-m3.toml b/providers/melious/models/minimax-m3.toml new file mode 100644 index 00000000000..961c17037e2 --- /dev/null +++ b/providers/melious/models/minimax-m3.toml @@ -0,0 +1,14 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/minimax-m3 and https://melious.ai/pricing +base_model = "minimax/MiniMax-M3" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.4614 +output = 2.307 +cache_read = 0.1154 + +[limit] +context = 1_000_000 diff --git a/providers/melious/models/mistral-medium-3.5-128b.toml b/providers/melious/models/mistral-medium-3.5-128b.toml new file mode 100644 index 00000000000..90b24266fe3 --- /dev/null +++ b/providers/melious/models/mistral-medium-3.5-128b.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/mistral-medium-3.5-128b and https://melious.ai/pricing +base_model = "mistral/mistral-medium-2604" +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 1.7302 +output = 5.7675 + +[limit] +context = 256_000 +output = 256_000 diff --git a/providers/melious/models/mistral-small-3.2-24b-instruct.toml b/providers/melious/models/mistral-small-3.2-24b-instruct.toml new file mode 100644 index 00000000000..26ba3e6a531 --- /dev/null +++ b/providers/melious/models/mistral-small-3.2-24b-instruct.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/mistral-small-3.2-24b-instruct and https://melious.ai/pricing +base_model = "mistral/mistral-small-2506" +attachment = true +structured_output = true + +[cost] +input = 0.1154 +output = 0.346 + +[limit] +context = 131_072 diff --git a/providers/melious/models/mistral-small-4-119b-instruct.toml b/providers/melious/models/mistral-small-4-119b-instruct.toml new file mode 100644 index 00000000000..d15d256361e --- /dev/null +++ b/providers/melious/models/mistral-small-4-119b-instruct.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/mistral-small-4-119b-instruct and https://melious.ai/pricing +base_model = "mistral/mistral-small-2603" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.173 +output = 0.6921 diff --git a/providers/melious/models/multilingual-e5-large-instruct.toml b/providers/melious/models/multilingual-e5-large-instruct.toml new file mode 100644 index 00000000000..3a45755b227 --- /dev/null +++ b/providers/melious/models/multilingual-e5-large-instruct.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/multilingual-e5-large-instruct and https://melious.ai/pricing +base_model = "intfloat/multilingual-e5-large-instruct" + +[cost] +input = 0.0346 +output = 0 diff --git a/providers/melious/models/multilingual-e5-large.toml b/providers/melious/models/multilingual-e5-large.toml new file mode 100644 index 00000000000..8ccac1ae842 --- /dev/null +++ b/providers/melious/models/multilingual-e5-large.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/multilingual-e5-large and https://melious.ai/pricing +base_model = "intfloat/multilingual-e5-large" + +[cost] +input = 0.0346 +output = 0 diff --git a/providers/melious/models/nemotron-3-nano-30b-a3b.toml b/providers/melious/models/nemotron-3-nano-30b-a3b.toml new file mode 100644 index 00000000000..da22e32b4a0 --- /dev/null +++ b/providers/melious/models/nemotron-3-nano-30b-a3b.toml @@ -0,0 +1,14 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/nemotron-3-nano-30b-a3b and https://melious.ai/pricing +base_model = "nvidia/nemotron-3-nano-30b-a3b" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.0692 +output = 0.2768 + +[limit] +context = 128_000 +output = 128_000 diff --git a/providers/melious/models/nemotron-3-super-120b-a12b.toml b/providers/melious/models/nemotron-3-super-120b-a12b.toml new file mode 100644 index 00000000000..0e5b762f51f --- /dev/null +++ b/providers/melious/models/nemotron-3-super-120b-a12b.toml @@ -0,0 +1,10 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/nemotron-3-super-120b-a12b-fp8 and https://melious.ai/pricing +base_model = "nvidia/nemotron-3-super-120b-a12b" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "low", "medium"] }] + +[cost] +input = 0.346 +output = 1.0381 diff --git a/providers/melious/models/paraphrase-multilingual-mpnet.toml b/providers/melious/models/paraphrase-multilingual-mpnet.toml new file mode 100644 index 00000000000..74c61bd16b9 --- /dev/null +++ b/providers/melious/models/paraphrase-multilingual-mpnet.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/paraphrase-multilingual-mpnet and https://melious.ai/pricing +base_model = "sentence-transformers/paraphrase-multilingual-mpnet-base-v2" + +[cost] +input = 0.0115 +output = 0 diff --git a/providers/melious/models/pixtral-12b-2409.toml b/providers/melious/models/pixtral-12b-2409.toml new file mode 100644 index 00000000000..16ffb39cb64 --- /dev/null +++ b/providers/melious/models/pixtral-12b-2409.toml @@ -0,0 +1,9 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/pixtral-12b-2409 and https://melious.ai/pricing +base_model = "mistral/pixtral-12b" +structured_output = true + +[cost] +input = 0.2307 +output = 0.2307 diff --git a/providers/melious/models/qwen2.5-vl-72b-instruct.toml b/providers/melious/models/qwen2.5-vl-72b-instruct.toml new file mode 100644 index 00000000000..f49cbba7cf0 --- /dev/null +++ b/providers/melious/models/qwen2.5-vl-72b-instruct.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen2.5-vl-72b-instruct and https://melious.ai/pricing +base_model = "alibaba/qwen2-5-vl-72b-instruct" +attachment = true +tool_call = false + +[cost] +input = 0.2884 +output = 0.8651 + +[limit] +context = 32_768 diff --git a/providers/melious/models/qwen3-235b-a22b-instruct.toml b/providers/melious/models/qwen3-235b-a22b-instruct.toml new file mode 100644 index 00000000000..02282dab437 --- /dev/null +++ b/providers/melious/models/qwen3-235b-a22b-instruct.toml @@ -0,0 +1,9 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-235b-a22b-instruct and https://melious.ai/pricing +base_model = "alibaba/qwen3-235b-a22b-instruct-2507" +structured_output = true + +[cost] +input = 0.2307 +output = 0.6921 diff --git a/providers/melious/models/qwen3-30b-a3b-instruct.toml b/providers/melious/models/qwen3-30b-a3b-instruct.toml new file mode 100644 index 00000000000..0edb9b3d62c --- /dev/null +++ b/providers/melious/models/qwen3-30b-a3b-instruct.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-30b-a3b-instruct and https://melious.ai/pricing +base_model = "alibaba/qwen3-30b-a3b-instruct-2507" + +[cost] +input = 0.1154 +output = 0.346 diff --git a/providers/melious/models/qwen3-32b.toml b/providers/melious/models/qwen3-32b.toml new file mode 100644 index 00000000000..981400f6011 --- /dev/null +++ b/providers/melious/models/qwen3-32b.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-32b and https://melious.ai/pricing +base_model = "alibaba/qwen3-32b" +structured_output = true +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.0923 +output = 0.2768 + +[limit] +context = 32_768 diff --git a/providers/melious/models/qwen3-coder-30b-a3b-instruct.toml b/providers/melious/models/qwen3-coder-30b-a3b-instruct.toml new file mode 100644 index 00000000000..02e90d3914d --- /dev/null +++ b/providers/melious/models/qwen3-coder-30b-a3b-instruct.toml @@ -0,0 +1,9 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-coder-30b-a3b-instruct and https://melious.ai/pricing +base_model = "alibaba/qwen3-coder-30b-a3b-instruct" +structured_output = true + +[cost] +input = 0.0692 +output = 0.2768 diff --git a/providers/melious/models/qwen3-coder-next.toml b/providers/melious/models/qwen3-coder-next.toml new file mode 100644 index 00000000000..8d9a6d23110 --- /dev/null +++ b/providers/melious/models/qwen3-coder-next.toml @@ -0,0 +1,12 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-coder-next and https://melious.ai/pricing +base_model = "alibaba/qwen3-coder-next" +structured_output = false + +[cost] +input = 0.1961 +output = 1.0381 + +[limit] +context = 131_072 diff --git a/providers/melious/models/qwen3-embedding-8b.toml b/providers/melious/models/qwen3-embedding-8b.toml new file mode 100644 index 00000000000..71349f263dd --- /dev/null +++ b/providers/melious/models/qwen3-embedding-8b.toml @@ -0,0 +1,11 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-embedding-8b and https://melious.ai/pricing +base_model = "alibaba/qwen3-embedding-8b" + +[cost] +input = 0.0115 +output = 0 + +[limit] +context = 41_000 diff --git a/providers/melious/models/qwen3-next-80b-a3b-thinking.toml b/providers/melious/models/qwen3-next-80b-a3b-thinking.toml new file mode 100644 index 00000000000..5408f631075 --- /dev/null +++ b/providers/melious/models/qwen3-next-80b-a3b-thinking.toml @@ -0,0 +1,12 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-next-80b-a3b-thinking and https://melious.ai/pricing +base_model = "alibaba/qwen3-next-80b-a3b-thinking" +reasoning_options = [] + +[cost] +input = 0.173 +output = 1.3842 + +[limit] +context = 262_144 diff --git a/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml b/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml new file mode 100644 index 00000000000..d6cc656fe84 --- /dev/null +++ b/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml @@ -0,0 +1,14 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3-vl-235b-a22b-instruct and https://melious.ai/pricing +base_model = "alibaba/qwen3-vl-235b-a22b-instruct" + +[cost] +input = 0.2307 +output = 2.0763 + +[limit] +context = 131_000 + +[modalities] +input = ["text", "image"] diff --git a/providers/melious/models/qwen3.5-122b-a10b.toml b/providers/melious/models/qwen3.5-122b-a10b.toml new file mode 100644 index 00000000000..03a4b0783a6 --- /dev/null +++ b/providers/melious/models/qwen3.5-122b-a10b.toml @@ -0,0 +1,15 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3.5-122b-a10b and https://melious.ai/pricing +base_model = "alibaba/qwen3.5-122b-a10b" +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.2538 +output = 2.2839 + +[limit] +context = 240_000 + +[modalities] +input = ["text", "image"] diff --git a/providers/melious/models/qwen3.5-397b-a17b.toml b/providers/melious/models/qwen3.5-397b-a17b.toml new file mode 100644 index 00000000000..6395c58b4b7 --- /dev/null +++ b/providers/melious/models/qwen3.5-397b-a17b.toml @@ -0,0 +1,15 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3.5-397b-a17b and https://melious.ai/pricing +base_model = "alibaba/qwen3.5-397b-a17b" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.6921 +output = 4.1526 + +[limit] +context = 256_000 + +[modalities] +input = ["text", "image"] diff --git a/providers/melious/models/qwen3.5-9b.toml b/providers/melious/models/qwen3.5-9b.toml new file mode 100644 index 00000000000..aee2540467c --- /dev/null +++ b/providers/melious/models/qwen3.5-9b.toml @@ -0,0 +1,13 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3.5-9b and https://melious.ai/pricing +base_model = "alibaba/qwen3.5-9b" +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.173 +output = 0.2307 +cache_read = 0.0461 + +[modalities] +input = ["text", "image"] diff --git a/providers/melious/models/qwen3.6-27b.toml b/providers/melious/models/qwen3.6-27b.toml new file mode 100644 index 00000000000..eeb8ade6eee --- /dev/null +++ b/providers/melious/models/qwen3.6-27b.toml @@ -0,0 +1,12 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3.6-27b and https://melious.ai/pricing +base_model = "alibaba/qwen3.6-27b" +reasoning_options = [{ type = "effort", values = ["none", "high"] }] + +[cost] +input = 0.2307 +output = 1.3842 + +[modalities] +input = ["text", "image", "video"] diff --git a/providers/melious/models/qwen3.6-35b-a3b.toml b/providers/melious/models/qwen3.6-35b-a3b.toml new file mode 100644 index 00000000000..e991b92e549 --- /dev/null +++ b/providers/melious/models/qwen3.6-35b-a3b.toml @@ -0,0 +1,15 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/qwen3.6-35b-a3b and https://melious.ai/pricing +base_model = "alibaba/qwen3.6-35b-a3b" +reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] + +[cost] +input = 0.173 +output = 0.5767 + +[limit] +context = 256_000 + +[modalities] +input = ["text", "image"] diff --git a/providers/melious/models/voxtral-small-24b-2507.toml b/providers/melious/models/voxtral-small-24b-2507.toml new file mode 100644 index 00000000000..13a2514d092 --- /dev/null +++ b/providers/melious/models/voxtral-small-24b-2507.toml @@ -0,0 +1,8 @@ +# Cost: converted from Melious's EUR list price for the default `balanced` +# route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). +# Sources: https://melious.ai/hub/models/voxtral-small-24b-2507 and https://melious.ai/pricing +base_model = "mistral/voxtral-small-2507" + +[cost] +input = 0.173 +output = 0.4037 diff --git a/providers/melious/provider.toml b/providers/melious/provider.toml new file mode 100644 index 00000000000..718bdb84de4 --- /dev/null +++ b/providers/melious/provider.toml @@ -0,0 +1,10 @@ +# Melious AI GmbH relays open-weight models on European infrastructure. Raw HTTP is +# POST `/v1/chat/completions`; the OpenAI SDK works by pointing `base_url` at the API. +# Routing flavours are selected with a model-ID suffix (`:balanced`, `:speed`, +# `:price`, `:eco`, `:batch`). A bare model ID resolves to `balanced`, which is what +# the entries here describe. https://melious.ai/docs/concepts/routing +name = "Melious" +npm = "@ai-sdk/openai-compatible" +api = "https://api.melious.ai/v1" +env = ["MELIOUS_API_KEY"] +doc = "https://melious.ai/docs/get-started/openai-sdk" From f05c753c6ab467c26f2af2f69b55e559845a36f4 Mon Sep 17 00:00:00 2001 From: Peter van der Spek Date: Tue, 18 Aug 2026 15:20:21 +0200 Subject: [PATCH 2/5] fix(melious): author reasoning_options from the documented wire format The API reference for this host documents exactly one reasoning control: `reasoning_effort` accepting low|medium|high for reasoning models, ignored by non-reasoning models. There is no thinking/toggle field and no reasoning budget field, so the previous per-model effort enums invented values this API does not accept (`none`, `max`) and narrowed others without evidence. - All 34 reasoning models now use the documented low|medium|high enum. - provider.toml records the wire field and the reference URL. - Drop the context overrides on bge-large-en-v1.5 and qwen3-embedding-8b: the hub reports 8192 and 41000, above the models' architectural maxima of 512 and 32768, so the lab limits are kept. Ref: https://melious.ai/docs/reference/chat-completions Co-Authored-By: Claude Opus 5 --- providers/melious/models/bge-large-en-v1.5.toml | 4 ++-- providers/melious/models/deepseek-r1-0528.toml | 3 +-- providers/melious/models/deepseek-v3.1.toml | 2 +- providers/melious/models/deepseek-v3.2.toml | 2 +- providers/melious/models/deepseek-v4-flash-0731.toml | 3 +-- providers/melious/models/deepseek-v4-flash.toml | 3 +-- providers/melious/models/deepseek-v4-pro.toml | 3 +-- providers/melious/models/gemma-4-26b-a4b.toml | 2 +- providers/melious/models/gemma-4-31b.toml | 2 +- providers/melious/models/glm-4.7.toml | 2 +- providers/melious/models/glm-5.1.toml | 2 +- providers/melious/models/glm-5.2.toml | 2 +- providers/melious/models/glm-5.toml | 2 +- providers/melious/models/hermes-4-405b.toml | 3 +-- providers/melious/models/hermes-4-70b.toml | 3 +-- providers/melious/models/kimi-k2.5.toml | 2 +- providers/melious/models/kimi-k2.6.toml | 2 +- providers/melious/models/kimi-k3.toml | 2 +- providers/melious/models/minimax-m2.5.toml | 2 +- providers/melious/models/minimax-m2.7.toml | 2 +- providers/melious/models/minimax-m3.toml | 2 +- providers/melious/models/mistral-medium-3.5-128b.toml | 2 +- .../melious/models/mistral-small-4-119b-instruct.toml | 2 +- providers/melious/models/nemotron-3-nano-30b-a3b.toml | 2 +- providers/melious/models/nemotron-3-super-120b-a12b.toml | 2 +- providers/melious/models/qwen3-32b.toml | 2 +- providers/melious/models/qwen3-embedding-8b.toml | 4 ++-- providers/melious/models/qwen3-next-80b-a3b-thinking.toml | 2 +- providers/melious/models/qwen3.5-122b-a10b.toml | 2 +- providers/melious/models/qwen3.5-397b-a17b.toml | 2 +- providers/melious/models/qwen3.5-9b.toml | 2 +- providers/melious/models/qwen3.6-27b.toml | 2 +- providers/melious/models/qwen3.6-35b-a3b.toml | 2 +- providers/melious/provider.toml | 8 +++++++- 34 files changed, 42 insertions(+), 42 deletions(-) diff --git a/providers/melious/models/bge-large-en-v1.5.toml b/providers/melious/models/bge-large-en-v1.5.toml index 103c5b19f12..4c33857c729 100644 --- a/providers/melious/models/bge-large-en-v1.5.toml +++ b/providers/melious/models/bge-large-en-v1.5.toml @@ -1,11 +1,11 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/bge-large-en-v1.5 and https://melious.ai/pricing +# Limit: the hub reports a 8192-token context for this ID, but BGE Large EN v1.5 is +# hard-capped at 512 positions, so the lab limit is kept rather than overridden. base_model = "baai/bge-large-en-v1.5" [cost] input = 0.0231 output = 0 -[limit] -context = 8192 diff --git a/providers/melious/models/deepseek-r1-0528.toml b/providers/melious/models/deepseek-r1-0528.toml index 8d34fd98f34..abf7abb26e8 100644 --- a/providers/melious/models/deepseek-r1-0528.toml +++ b/providers/melious/models/deepseek-r1-0528.toml @@ -1,11 +1,10 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/deepseek-r1-0528 and https://melious.ai/pricing -# R1 always reasons and exposes no effort control; served as the 0528 checkpoint. base_model = "deepseek/deepseek-r1" name = "DeepSeek R1 0528" structured_output = true -reasoning_options = [] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.7498 diff --git a/providers/melious/models/deepseek-v3.1.toml b/providers/melious/models/deepseek-v3.1.toml index 3384b5e1bdb..6b7c9aee12b 100644 --- a/providers/melious/models/deepseek-v3.1.toml +++ b/providers/melious/models/deepseek-v3.1.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/deepseek-v3.1 and https://melious.ai/pricing base_model = "deepseek/deepseek-v3.1" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.2307 diff --git a/providers/melious/models/deepseek-v3.2.toml b/providers/melious/models/deepseek-v3.2.toml index 260d66b711e..79581eb1add 100644 --- a/providers/melious/models/deepseek-v3.2.toml +++ b/providers/melious/models/deepseek-v3.2.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/deepseek-v3.2 and https://melious.ai/pricing base_model = "deepseek/deepseek-v3.2" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.346 diff --git a/providers/melious/models/deepseek-v4-flash-0731.toml b/providers/melious/models/deepseek-v4-flash-0731.toml index d1dd5455ea4..2e4cea8e40f 100644 --- a/providers/melious/models/deepseek-v4-flash-0731.toml +++ b/providers/melious/models/deepseek-v4-flash-0731.toml @@ -1,9 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/deepseek-v4-flash-0731 and https://melious.ai/pricing -# Effort: DeepSeek V4 collapses low/medium onto high, so only high and max differ. base_model = "deepseek/deepseek-v4-flash-0731" -reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.2884 diff --git a/providers/melious/models/deepseek-v4-flash.toml b/providers/melious/models/deepseek-v4-flash.toml index 5a8e66bd40e..04512c96472 100644 --- a/providers/melious/models/deepseek-v4-flash.toml +++ b/providers/melious/models/deepseek-v4-flash.toml @@ -1,9 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/deepseek-v4-flash and https://melious.ai/pricing -# Effort: DeepSeek V4 collapses low/medium onto high, so only high and max differ. base_model = "deepseek/deepseek-v4-flash" -reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.173 diff --git a/providers/melious/models/deepseek-v4-pro.toml b/providers/melious/models/deepseek-v4-pro.toml index 0b47544b65a..ec810e5ba3a 100644 --- a/providers/melious/models/deepseek-v4-pro.toml +++ b/providers/melious/models/deepseek-v4-pro.toml @@ -1,9 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/deepseek-v4-pro and https://melious.ai/pricing -# Effort: DeepSeek V4 collapses low/medium onto high, so only high and max differ. base_model = "deepseek/deepseek-v4-pro" -reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 1.8456 diff --git a/providers/melious/models/gemma-4-26b-a4b.toml b/providers/melious/models/gemma-4-26b-a4b.toml index dafd7234515..64dcb0ae049 100644 --- a/providers/melious/models/gemma-4-26b-a4b.toml +++ b/providers/melious/models/gemma-4-26b-a4b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/gemma-4-26b-a4b and https://melious.ai/pricing base_model = "google/gemma-4-26b-a4b-it" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.1154 diff --git a/providers/melious/models/gemma-4-31b.toml b/providers/melious/models/gemma-4-31b.toml index dd115c5f951..4eccffef144 100644 --- a/providers/melious/models/gemma-4-31b.toml +++ b/providers/melious/models/gemma-4-31b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/gemma-4-31b and https://melious.ai/pricing base_model = "google/gemma-4-31b-it" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.1154 diff --git a/providers/melious/models/glm-4.7.toml b/providers/melious/models/glm-4.7.toml index 73e61f23916..c494fa736df 100644 --- a/providers/melious/models/glm-4.7.toml +++ b/providers/melious/models/glm-4.7.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/glm-4.7 and https://melious.ai/pricing base_model = "zhipuai/glm-4.7" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.8074 diff --git a/providers/melious/models/glm-5.1.toml b/providers/melious/models/glm-5.1.toml index 16feb6fa624..6f488a97208 100644 --- a/providers/melious/models/glm-5.1.toml +++ b/providers/melious/models/glm-5.1.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/glm-5.1 and https://melious.ai/pricing base_model = "zhipuai/glm-5.1" -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 1.4995 diff --git a/providers/melious/models/glm-5.2.toml b/providers/melious/models/glm-5.2.toml index ea757e61d8e..9f85e673065 100644 --- a/providers/melious/models/glm-5.2.toml +++ b/providers/melious/models/glm-5.2.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/glm-5.2 and https://melious.ai/pricing base_model = "zhipuai/glm-5.2" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 1.1535 diff --git a/providers/melious/models/glm-5.toml b/providers/melious/models/glm-5.toml index 192de4b0168..7a9d8021a80 100644 --- a/providers/melious/models/glm-5.toml +++ b/providers/melious/models/glm-5.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/glm-5 and https://melious.ai/pricing base_model = "zhipuai/glm-5" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 1.0958 diff --git a/providers/melious/models/hermes-4-405b.toml b/providers/melious/models/hermes-4-405b.toml index 7e6b3dac15e..2565db32d3f 100644 --- a/providers/melious/models/hermes-4-405b.toml +++ b/providers/melious/models/hermes-4-405b.toml @@ -1,9 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/hermes-4-405b and https://melious.ai/pricing -# Hermes 4 hybrid reasoning is steered by the system prompt, not a request field. base_model = "nousresearch/hermes-4-405b" -reasoning_options = [] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 1.0958 diff --git a/providers/melious/models/hermes-4-70b.toml b/providers/melious/models/hermes-4-70b.toml index f30144d603e..f02c70248c1 100644 --- a/providers/melious/models/hermes-4-70b.toml +++ b/providers/melious/models/hermes-4-70b.toml @@ -1,9 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/hermes-4-70b and https://melious.ai/pricing -# Hermes 4 hybrid reasoning is steered by the system prompt, not a request field. base_model = "nousresearch/hermes-4-70b" -reasoning_options = [] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.1384 diff --git a/providers/melious/models/kimi-k2.5.toml b/providers/melious/models/kimi-k2.5.toml index 41f9aae016f..535cb34ebcc 100644 --- a/providers/melious/models/kimi-k2.5.toml +++ b/providers/melious/models/kimi-k2.5.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/kimi-k2.5 and https://melious.ai/pricing base_model = "moonshotai/kimi-k2.5" -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.5767 diff --git a/providers/melious/models/kimi-k2.6.toml b/providers/melious/models/kimi-k2.6.toml index d5190059a7b..452e2dc296d 100644 --- a/providers/melious/models/kimi-k2.6.toml +++ b/providers/melious/models/kimi-k2.6.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/kimi-k2.6 and https://melious.ai/pricing base_model = "moonshotai/kimi-k2.6" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.8074 diff --git a/providers/melious/models/kimi-k3.toml b/providers/melious/models/kimi-k3.toml index 8bf9fd588dd..9ff8b694beb 100644 --- a/providers/melious/models/kimi-k3.toml +++ b/providers/melious/models/kimi-k3.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/kimi-k3 and https://melious.ai/pricing base_model = "moonshotai/kimi-k3" -reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 3.4605 diff --git a/providers/melious/models/minimax-m2.5.toml b/providers/melious/models/minimax-m2.5.toml index 915a15b3cf8..7fdb704720b 100644 --- a/providers/melious/models/minimax-m2.5.toml +++ b/providers/melious/models/minimax-m2.5.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/minimax-m2.5 and https://melious.ai/pricing base_model = "minimax/MiniMax-M2.5" structured_output = true -reasoning_options = [] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.2999 diff --git a/providers/melious/models/minimax-m2.7.toml b/providers/melious/models/minimax-m2.7.toml index 6fc55563ed8..4a34b1aad84 100644 --- a/providers/melious/models/minimax-m2.7.toml +++ b/providers/melious/models/minimax-m2.7.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/minimax-m2.7 and https://melious.ai/pricing base_model = "minimax/MiniMax-M2.7" structured_output = true -reasoning_options = [] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.6921 diff --git a/providers/melious/models/minimax-m3.toml b/providers/melious/models/minimax-m3.toml index 961c17037e2..9d5e0e59406 100644 --- a/providers/melious/models/minimax-m3.toml +++ b/providers/melious/models/minimax-m3.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/minimax-m3 and https://melious.ai/pricing base_model = "minimax/MiniMax-M3" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.4614 diff --git a/providers/melious/models/mistral-medium-3.5-128b.toml b/providers/melious/models/mistral-medium-3.5-128b.toml index 90b24266fe3..872a5e8071e 100644 --- a/providers/melious/models/mistral-medium-3.5-128b.toml +++ b/providers/melious/models/mistral-medium-3.5-128b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/mistral-medium-3.5-128b and https://melious.ai/pricing base_model = "mistral/mistral-medium-2604" -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 1.7302 diff --git a/providers/melious/models/mistral-small-4-119b-instruct.toml b/providers/melious/models/mistral-small-4-119b-instruct.toml index d15d256361e..4b0d05db665 100644 --- a/providers/melious/models/mistral-small-4-119b-instruct.toml +++ b/providers/melious/models/mistral-small-4-119b-instruct.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/mistral-small-4-119b-instruct and https://melious.ai/pricing base_model = "mistral/mistral-small-2603" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.173 diff --git a/providers/melious/models/nemotron-3-nano-30b-a3b.toml b/providers/melious/models/nemotron-3-nano-30b-a3b.toml index da22e32b4a0..6b9893f1a77 100644 --- a/providers/melious/models/nemotron-3-nano-30b-a3b.toml +++ b/providers/melious/models/nemotron-3-nano-30b-a3b.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/nemotron-3-nano-30b-a3b and https://melious.ai/pricing base_model = "nvidia/nemotron-3-nano-30b-a3b" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.0692 diff --git a/providers/melious/models/nemotron-3-super-120b-a12b.toml b/providers/melious/models/nemotron-3-super-120b-a12b.toml index 0e5b762f51f..155de489c04 100644 --- a/providers/melious/models/nemotron-3-super-120b-a12b.toml +++ b/providers/melious/models/nemotron-3-super-120b-a12b.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/nemotron-3-super-120b-a12b-fp8 and https://melious.ai/pricing base_model = "nvidia/nemotron-3-super-120b-a12b" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "low", "medium"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.346 diff --git a/providers/melious/models/qwen3-32b.toml b/providers/melious/models/qwen3-32b.toml index 981400f6011..20c66810b71 100644 --- a/providers/melious/models/qwen3-32b.toml +++ b/providers/melious/models/qwen3-32b.toml @@ -3,7 +3,7 @@ # Sources: https://melious.ai/hub/models/qwen3-32b and https://melious.ai/pricing base_model = "alibaba/qwen3-32b" structured_output = true -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.0923 diff --git a/providers/melious/models/qwen3-embedding-8b.toml b/providers/melious/models/qwen3-embedding-8b.toml index 71349f263dd..ac406602e5a 100644 --- a/providers/melious/models/qwen3-embedding-8b.toml +++ b/providers/melious/models/qwen3-embedding-8b.toml @@ -1,11 +1,11 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3-embedding-8b and https://melious.ai/pricing +# Limit: the hub reports a 41000-token context, above Qwen3-Embedding-8B's native +# 32768, so the lab limit is kept rather than overridden. base_model = "alibaba/qwen3-embedding-8b" [cost] input = 0.0115 output = 0 -[limit] -context = 41_000 diff --git a/providers/melious/models/qwen3-next-80b-a3b-thinking.toml b/providers/melious/models/qwen3-next-80b-a3b-thinking.toml index 5408f631075..467fd750f30 100644 --- a/providers/melious/models/qwen3-next-80b-a3b-thinking.toml +++ b/providers/melious/models/qwen3-next-80b-a3b-thinking.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3-next-80b-a3b-thinking and https://melious.ai/pricing base_model = "alibaba/qwen3-next-80b-a3b-thinking" -reasoning_options = [] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.173 diff --git a/providers/melious/models/qwen3.5-122b-a10b.toml b/providers/melious/models/qwen3.5-122b-a10b.toml index 03a4b0783a6..e9fc18267a6 100644 --- a/providers/melious/models/qwen3.5-122b-a10b.toml +++ b/providers/melious/models/qwen3.5-122b-a10b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3.5-122b-a10b and https://melious.ai/pricing base_model = "alibaba/qwen3.5-122b-a10b" -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.2538 diff --git a/providers/melious/models/qwen3.5-397b-a17b.toml b/providers/melious/models/qwen3.5-397b-a17b.toml index 6395c58b4b7..2f4512defc1 100644 --- a/providers/melious/models/qwen3.5-397b-a17b.toml +++ b/providers/melious/models/qwen3.5-397b-a17b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3.5-397b-a17b and https://melious.ai/pricing base_model = "alibaba/qwen3.5-397b-a17b" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.6921 diff --git a/providers/melious/models/qwen3.5-9b.toml b/providers/melious/models/qwen3.5-9b.toml index aee2540467c..750a09a3658 100644 --- a/providers/melious/models/qwen3.5-9b.toml +++ b/providers/melious/models/qwen3.5-9b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3.5-9b and https://melious.ai/pricing base_model = "alibaba/qwen3.5-9b" -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.173 diff --git a/providers/melious/models/qwen3.6-27b.toml b/providers/melious/models/qwen3.6-27b.toml index eeb8ade6eee..5dbdb445fae 100644 --- a/providers/melious/models/qwen3.6-27b.toml +++ b/providers/melious/models/qwen3.6-27b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3.6-27b and https://melious.ai/pricing base_model = "alibaba/qwen3.6-27b" -reasoning_options = [{ type = "effort", values = ["none", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.2307 diff --git a/providers/melious/models/qwen3.6-35b-a3b.toml b/providers/melious/models/qwen3.6-35b-a3b.toml index e991b92e549..89557d88497 100644 --- a/providers/melious/models/qwen3.6-35b-a3b.toml +++ b/providers/melious/models/qwen3.6-35b-a3b.toml @@ -2,7 +2,7 @@ # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3.6-35b-a3b and https://melious.ai/pricing base_model = "alibaba/qwen3.6-35b-a3b" -reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }] +reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] [cost] input = 0.173 diff --git a/providers/melious/provider.toml b/providers/melious/provider.toml index 718bdb84de4..4e2c5bc1644 100644 --- a/providers/melious/provider.toml +++ b/providers/melious/provider.toml @@ -1,8 +1,14 @@ # Melious AI GmbH relays open-weight models on European infrastructure. Raw HTTP is # POST `/v1/chat/completions`; the OpenAI SDK works by pointing `base_url` at the API. +# Reasoning: this host flattens every upstream thinking control into one OpenAI-style +# `reasoning_effort` field accepting exactly low|medium|high for reasoning models, and +# ignores it for non-reasoning models. There is no thinking/toggle field and no +# reasoning budget field, so no model here uses `toggle` or `budget_tokens`. +# https://melious.ai/docs/reference/chat-completions # Routing flavours are selected with a model-ID suffix (`:balanced`, `:speed`, # `:price`, `:eco`, `:batch`). A bare model ID resolves to `balanced`, which is what -# the entries here describe. https://melious.ai/docs/concepts/routing +# the entries here describe. `preset` biases routing, not reasoning depth. +# https://melious.ai/docs/concepts/routing name = "Melious" npm = "@ai-sdk/openai-compatible" api = "https://api.melious.ai/v1" From d390471f259fd5f6df5b42625ca66b7dad3bca16 Mon Sep 17 00:00:00 2001 From: Peter van der Spek Date: Tue, 18 Aug 2026 15:25:07 +0200 Subject: [PATCH 3/5] fix(melious): drop overrides made redundant by the dev merge dev's canonical models/alibaba/qwen3-vl-235b-a22b-instruct.toml (taken over mine in the merge) states a 131072 context and text+image modalities, and 06f44f509 set meta/llama-3.1-8b-instruct context to 128000. Both now match what Melious serves, so the provider entries no longer need to restate them. Co-Authored-By: Claude Opus 5 --- providers/melious/models/llama-3.1-8b-instruct.toml | 2 -- providers/melious/models/qwen3-vl-235b-a22b-instruct.toml | 4 ---- 2 files changed, 6 deletions(-) diff --git a/providers/melious/models/llama-3.1-8b-instruct.toml b/providers/melious/models/llama-3.1-8b-instruct.toml index de5364e122d..21a2feed889 100644 --- a/providers/melious/models/llama-3.1-8b-instruct.toml +++ b/providers/melious/models/llama-3.1-8b-instruct.toml @@ -7,5 +7,3 @@ base_model = "meta/llama-3.1-8b-instruct" input = 0.173 output = 0.173 -[limit] -context = 128_000 diff --git a/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml b/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml index d6cc656fe84..620efd8b7fb 100644 --- a/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml +++ b/providers/melious/models/qwen3-vl-235b-a22b-instruct.toml @@ -7,8 +7,4 @@ base_model = "alibaba/qwen3-vl-235b-a22b-instruct" input = 0.2307 output = 2.0763 -[limit] -context = 131_000 -[modalities] -input = ["text", "image"] From 56a1aeac1c71038a74255e8130c275b808467d34 Mon Sep 17 00:00:00 2001 From: Peter van der Spek Date: Wed, 19 Aug 2026 10:07:25 +0200 Subject: [PATCH 4/5] fix(melious): use [] where no upstream reasoning dial exists Melious only forwards `reasoning_effort` (low|medium|high), so authoring that enum for models with no upstream effort concept overstated caller control. Set [] for the always-on reasoners and the prompt-steered hybrids, each with a leading comment citing the baseline: - deepseek-r1-0528, minimax-m2.5, minimax-m2.7, kimi-k2.7-code: the first-party lab host authors [] (deepseek-reasoner, MiniMax-M2.5/M2.7, kimi-k2.7-code). - qwen3-next-80b-a3b-thinking: Alibaba exposes only a reasoning budget, which Melious has no field for; OpenRouter authors []. - hermes-4-405b, hermes-4-70b: thinking is system-prompt steered; OpenRouter models it as a toggle, which Melious has no wire field for. The other 27 keep low|medium|high: their underlying models have a real on/off or graded thinking control, and effort is the only dial this host exposes. Also revert the [[links]] addition to models/meta/llama-3.1-8b-instruct.toml so this PR no longer modifies shared lab metadata that already exists on dev. Co-Authored-By: Claude Opus 5 --- models/meta/llama-3.1-8b-instruct.toml | 5 ----- providers/melious/models/deepseek-r1-0528.toml | 5 ++++- providers/melious/models/hermes-4-405b.toml | 4 +++- providers/melious/models/hermes-4-70b.toml | 4 +++- providers/melious/models/kimi-k2.7-code.toml | 4 +++- providers/melious/models/minimax-m2.5.toml | 4 +++- providers/melious/models/minimax-m2.7.toml | 4 +++- providers/melious/models/qwen3-next-80b-a3b-thinking.toml | 5 ++++- 8 files changed, 23 insertions(+), 12 deletions(-) diff --git a/models/meta/llama-3.1-8b-instruct.toml b/models/meta/llama-3.1-8b-instruct.toml index 7117fae282c..ff2584e0bab 100644 --- a/models/meta/llama-3.1-8b-instruct.toml +++ b/models/meta/llama-3.1-8b-instruct.toml @@ -21,8 +21,3 @@ output = ["text"] [[weights]] label = "Hugging Face" url = "https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct" - -[[links]] -label = "Model card" -url = "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md" -type = "model_card" diff --git a/providers/melious/models/deepseek-r1-0528.toml b/providers/melious/models/deepseek-r1-0528.toml index abf7abb26e8..6fec88c45be 100644 --- a/providers/melious/models/deepseek-r1-0528.toml +++ b/providers/melious/models/deepseek-r1-0528.toml @@ -1,10 +1,13 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/deepseek-r1-0528 and https://melious.ai/pricing +# Reasoning: R1 always reasons and exposes no effort dial. First-party +# providers/deepseek/models/deepseek-reasoner.toml and OpenRouter's r1-0528 both +# author [], so there is no caller-side control to forward here. base_model = "deepseek/deepseek-r1" name = "DeepSeek R1 0528" structured_output = true -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 0.7498 diff --git a/providers/melious/models/hermes-4-405b.toml b/providers/melious/models/hermes-4-405b.toml index 2565db32d3f..729c50cf522 100644 --- a/providers/melious/models/hermes-4-405b.toml +++ b/providers/melious/models/hermes-4-405b.toml @@ -1,8 +1,10 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/hermes-4-405b and https://melious.ai/pricing +# Reasoning: Hermes 4's hybrid thinking is steered by the system prompt. OpenRouter +# models it as a toggle, which Melious has no wire field for, so no caller control. base_model = "nousresearch/hermes-4-405b" -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 1.0958 diff --git a/providers/melious/models/hermes-4-70b.toml b/providers/melious/models/hermes-4-70b.toml index f02c70248c1..ce4bcd94581 100644 --- a/providers/melious/models/hermes-4-70b.toml +++ b/providers/melious/models/hermes-4-70b.toml @@ -1,8 +1,10 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/hermes-4-70b and https://melious.ai/pricing +# Reasoning: Hermes 4's hybrid thinking is steered by the system prompt. OpenRouter +# models it as a toggle, which Melious has no wire field for, so no caller control. base_model = "nousresearch/hermes-4-70b" -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 0.1384 diff --git a/providers/melious/models/kimi-k2.7-code.toml b/providers/melious/models/kimi-k2.7-code.toml index 827555cacf2..70e12784fac 100644 --- a/providers/melious/models/kimi-k2.7-code.toml +++ b/providers/melious/models/kimi-k2.7-code.toml @@ -1,8 +1,10 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/kimi-k2.7-code and https://melious.ai/pricing +# Reasoning: always-on. First-party providers/moonshotai/models/kimi-k2.7-code.toml +# authors [], so there is no caller-side control to forward here. base_model = "moonshotai/kimi-k2.7-code" -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 0.8074 diff --git a/providers/melious/models/minimax-m2.5.toml b/providers/melious/models/minimax-m2.5.toml index 7fdb704720b..45e8793dd51 100644 --- a/providers/melious/models/minimax-m2.5.toml +++ b/providers/melious/models/minimax-m2.5.toml @@ -1,9 +1,11 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/minimax-m2.5 and https://melious.ai/pricing +# Reasoning: always-on. First-party providers/minimax/models/MiniMax-M2.5.toml +# authors [], so there is no caller-side control to forward here. base_model = "minimax/MiniMax-M2.5" structured_output = true -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 0.2999 diff --git a/providers/melious/models/minimax-m2.7.toml b/providers/melious/models/minimax-m2.7.toml index 4a34b1aad84..f926d4cbefe 100644 --- a/providers/melious/models/minimax-m2.7.toml +++ b/providers/melious/models/minimax-m2.7.toml @@ -1,9 +1,11 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/minimax-m2.7 and https://melious.ai/pricing +# Reasoning: always-on. First-party providers/minimax/models/MiniMax-M2.7.toml +# authors [], so there is no caller-side control to forward here. base_model = "minimax/MiniMax-M2.7" structured_output = true -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 0.6921 diff --git a/providers/melious/models/qwen3-next-80b-a3b-thinking.toml b/providers/melious/models/qwen3-next-80b-a3b-thinking.toml index 467fd750f30..157b054b9e0 100644 --- a/providers/melious/models/qwen3-next-80b-a3b-thinking.toml +++ b/providers/melious/models/qwen3-next-80b-a3b-thinking.toml @@ -1,8 +1,11 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3-next-80b-a3b-thinking and https://melious.ai/pricing +# Reasoning: dedicated thinking ID. Alibaba's own API exposes only a reasoning +# budget (no effort grades) and OpenRouter authors []; Melious has no budget field, +# so there is no caller-side control to forward here. base_model = "alibaba/qwen3-next-80b-a3b-thinking" -reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] +reasoning_options = [] [cost] input = 0.173 From 31974445f9c1982cc523d7b8aa4675eae82ac89f Mon Sep 17 00:00:00 2001 From: Peter van der Spek Date: Wed, 19 Aug 2026 12:47:51 +0200 Subject: [PATCH 5/5] fix(melious): take limit.output from the hub's published max_tokens The hub payload's `parameters` block carries a `max_tokens` alongside `context_window`; the earlier scrape only read the latter, so limit.output was a mechanical clamp to the served context instead of the host's real cap. - devstral-2-123b-instruct-2512: output 200_000 -> 8_192 - nemotron-3-nano-30b-a3b: output 128_000 -> 8_192 - kimi-k2.5, mistral-medium-3.5-128b: the hub publishes no max_tokens for these, so the cap at the served context stays, now with a comment saying so. - qwen3-coder-next: cite capabilities.structured_output = false behind the flip. Cross-checked the whole provider against a fresh scrape: all 18 structured_output overrides and all 24 context overrides match the hub exactly. Co-Authored-By: Claude Opus 5 --- providers/melious/models/devstral-2-123b-instruct-2512.toml | 4 +++- providers/melious/models/kimi-k2.5.toml | 2 ++ providers/melious/models/mistral-medium-3.5-128b.toml | 2 ++ providers/melious/models/nemotron-3-nano-30b-a3b.toml | 4 +++- providers/melious/models/qwen3-coder-next.toml | 2 ++ 5 files changed, 12 insertions(+), 2 deletions(-) diff --git a/providers/melious/models/devstral-2-123b-instruct-2512.toml b/providers/melious/models/devstral-2-123b-instruct-2512.toml index 9b6e41493e2..80feb7becba 100644 --- a/providers/melious/models/devstral-2-123b-instruct-2512.toml +++ b/providers/melious/models/devstral-2-123b-instruct-2512.toml @@ -1,6 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/devstral-2-123b-instruct-2512 and https://melious.ai/pricing +# Limit: context and max output are the hub's served values for this deployment +# (parameters.context_window = 200000, parameters.max_tokens = 8192). base_model = "mistral/devstral-2512" structured_output = true @@ -10,4 +12,4 @@ output = 2.307 [limit] context = 200_000 -output = 200_000 +output = 8_192 diff --git a/providers/melious/models/kimi-k2.5.toml b/providers/melious/models/kimi-k2.5.toml index 535cb34ebcc..800028de0a7 100644 --- a/providers/melious/models/kimi-k2.5.toml +++ b/providers/melious/models/kimi-k2.5.toml @@ -1,6 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/kimi-k2.5 and https://melious.ai/pricing +# Limit: the hub states a 256000 served context but publishes no parameters.max_tokens, +# so output is capped at the served context rather than inheriting the lab's larger value. base_model = "moonshotai/kimi-k2.5" reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] diff --git a/providers/melious/models/mistral-medium-3.5-128b.toml b/providers/melious/models/mistral-medium-3.5-128b.toml index 872a5e8071e..9a9121c17ab 100644 --- a/providers/melious/models/mistral-medium-3.5-128b.toml +++ b/providers/melious/models/mistral-medium-3.5-128b.toml @@ -1,6 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/mistral-medium-3.5-128b and https://melious.ai/pricing +# Limit: the hub states a 256000 served context but publishes no parameters.max_tokens, +# so output is capped at the served context rather than inheriting the lab's larger value. base_model = "mistral/mistral-medium-2604" reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] diff --git a/providers/melious/models/nemotron-3-nano-30b-a3b.toml b/providers/melious/models/nemotron-3-nano-30b-a3b.toml index 6b9893f1a77..729a1f9b76a 100644 --- a/providers/melious/models/nemotron-3-nano-30b-a3b.toml +++ b/providers/melious/models/nemotron-3-nano-30b-a3b.toml @@ -1,6 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/nemotron-3-nano-30b-a3b and https://melious.ai/pricing +# Limit: context and max output are the hub's served values for this deployment +# (parameters.context_window = 128000, parameters.max_tokens = 8192). base_model = "nvidia/nemotron-3-nano-30b-a3b" structured_output = true reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }] @@ -11,4 +13,4 @@ output = 0.2768 [limit] context = 128_000 -output = 128_000 +output = 8_192 diff --git a/providers/melious/models/qwen3-coder-next.toml b/providers/melious/models/qwen3-coder-next.toml index 8d9a6d23110..895a4757a4f 100644 --- a/providers/melious/models/qwen3-coder-next.toml +++ b/providers/melious/models/qwen3-coder-next.toml @@ -1,6 +1,8 @@ # Cost: converted from Melious's EUR list price for the default `balanced` # route at 1.1535 USD/EUR (ECB euro reference rate, 2026-08-07). # Sources: https://melious.ai/hub/models/qwen3-coder-next and https://melious.ai/pricing +# structured_output: the hub reports capabilities.structured_output = false for this ID, +# unlike the lab entry. base_model = "alibaba/qwen3-coder-next" structured_output = false