[model-inventory] Model alias inventory update - 2026-07-28

[github-actions]

Summary

Model inventory analysis completed for 2026-07-28. All major model families are covered by existing aliases. Key findings: new Anthropic models (claude-opus-4.7/4.8/5, claude-sonnet-5) are fully covered by existing wildcard aliases. One notable gap: trajectory-compaction (Fireworks) has no alias. Pricing multiplier data is unavailable from Copilot API (all billing.multiplier fields return null); docs pricing table was successfully extracted via Playwright.

• Providers queried: Anthropic (10 models), Gemini (50 models), OpenAI (219 models), Copilot/reflect (37 models)
• Total models found: 316 across providers
• Proposed alias changes: 1 (add trajectory-compaction alias)
• Pricing gaps found: 14 Copilot models not in models.json (mostly legacy gpt-4/gpt-4o variants); billing multipliers unavailable from API

Provider Model Counts

Provider	Models Available	Status
anthropic	10	✅ ok
gemini	50	✅ ok
openai	219	✅ ok (219 includes internal/experimental IDs)
copilot (reflect)	37	✅ ok

Raw API Fields Discovered

Anthropic raw fields (/tmp/.../artifacts/anthropic-models/raw.json):

• id, display_name, created_at, type, max_input_tokens, max_tokens
• capabilities: rich nested object with batch, citations, code_execution, context_management, effort (including xhigh, max), image_input, pdf_input, structured_outputs, thinking
• No billing/pricing fields exposed
• Useful: effort.supported flags indicate extended reasoning support per model

Gemini raw fields (/tmp/.../artifacts/gemini-models/raw.json):

• id, display_name, description, version
• input_token_limit, output_token_limit — good proxy for pricing tier
• supported_generation_methods — e.g. generateContent, countTokens
• No pricing/billing fields; token limits serve as a complexity proxy

OpenAI raw fields (/tmp/.../artifacts/openai-models/raw.json):

• id, object, created, owned_by
• Minimal metadata; no pricing/context window fields in public models list
• owned_by: "openai", "system", "openai-internal" help filter internal/experimental models

Copilot (reflect endpoint) (/tmp/.../reflect.json):

• id, name, vendor, version, object, preview
• capabilities.limits: max_context_window_tokens, max_output_tokens, max_prompt_tokens, max_non_streaming_output_tokens
• capabilities.supports: streaming, tool_calls, parallel_tool_calls, vision, structured_outputs, reasoning_effort, adaptive_thinking
• capabilities.type: "chat" or "embeddings"
• supported_endpoints: e.g. ["/v1/messages", "/chat/completions"] or ["/responses", "ws:/responses"]
• billing.multiplier: present in schema but all returned null in this run
• model_picker_category: "powerful", "lightweight", "versatile"

models.json Pricing Analysis

Billing multipliers from Copilot API returned null for all models. Docs pricing table (from https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing) was successfully extracted via Playwright and covers OpenAI GPT-5.x family only.

Docs pricing table summary (per 1M tokens):

Model	Category	Input	Output
GPT-5 mini	Lightweight	$0.25	$2.00
GPT-5.3-Codex	Powerful	$1.75	$14.00
GPT-5.4 (≤272K)	Versatile	$2.50	$15.00
GPT-5.4 mini	Lightweight	$0.75	$4.50
GPT-5.5 (≤272K)	Powerful	$5.00	$30.00
GPT-5.6 Luna (≤200K)	Lightweight	$1.00	$6.00
GPT-5.6 Sol (≤272K)	Powerful	$5.00	$30.00
GPT-5.6 Terra (≤272K)	Versatile	$2.50	$15.00

Note: gpt-5.4-nano appears in the pricing table ($0.20 input / $1.25 output) but is on the intentionally excluded list per task instructions.

Missing from models.json (Copilot live models not in pkg/cli/data/models.json)

These Copilot live models are absent from models.json — most are legacy Azure-hosted models:

Model ID	Provider	Notes
gpt-3.5-turbo	copilot	Legacy Azure OpenAI
gpt-3.5-turbo-0613	copilot	Legacy Azure OpenAI
gpt-4	copilot	Legacy Azure OpenAI
gpt-4-0613	copilot	Legacy Azure OpenAI
gpt-4-o-preview	copilot	Legacy Azure OpenAI
gpt-4.1-2025-04-14	copilot	Azure OpenAI versioned
gpt-4o-2024-05-13	copilot	Azure OpenAI versioned
gpt-4o-2024-08-06	copilot	Azure OpenAI versioned
gpt-4o-2024-11-20	copilot	Azure OpenAI versioned
gpt-4o-mini-2024-07-18	copilot	Azure OpenAI versioned
text-embedding-3-small	copilot	Embeddings model
text-embedding-3-small-inference	copilot	Embeddings model
text-embedding-ada-002	copilot	Embeddings model
trajectory-compaction	copilot	Fireworks vendor; no alias

The first 13 are legacy/versioned Azure-hosted models. They may be candidates for models.json addition but are lower priority. trajectory-compaction is a specialized Fireworks model and warrants an alias.

Historical entries not currently returned (in models.json but absent from live APIs)

These remain as historical records and should not be removed without explicit human decision:

Model ID	Last seen notes
claude-fable-5	Anthropic family; not in Copilot live
claude-sonnet-4	Superseded by claude-sonnet-4.5+
gemini-3.1-pro-preview	Preview; may have graduated/retired
gemini-3.5-flash	Not in Copilot live (is in Google API)
gpt-5.2	Superseded by gpt-5.3+
gpt-5.2-codex	Superseded
gpt-5.4-nano	Deprecated Copilot ID (excluded)
kimi-k2.7-code	Not in current Copilot live
raptor-mini	Not in current Copilot live

Note: gemini-3.5-flash is returned by the Google Gemini API directly but is absent from the Copilot reflect endpoint.

Inferred vs stored pricing discrepancies

No meaningful discrepancies found for existing models.json entries — billing multipliers from Copilot API returned null for all models, preventing direct comparison. Docs pricing table does not cover Anthropic or Gemini models.

Proposed Alias Updates

1. Add trajectory-compaction alias

What: New alias trajectory-compaction pointing to the Fireworks-hosted trajectory compaction model on Copilot.

Why: trajectory-compaction is a live Copilot model (type: chat, vendor: Fireworks, endpoint: /chat/completions) with no current alias coverage. It's a specialized model likely used for conversation/context compaction tasks.

Syntax:

{
  "trajectory-compaction": [
    "copilot/trajectory-compaction*"
  ]
}

2. Add gemini-3.5-flash-lite alias

What: New alias for the Gemini 3.5 Flash Lite model, which has a different naming pattern than gemini-flash-lite (copilot/gemini-*flash*lite* does match it on Copilot if routed, but it's only in the Google API currently).

Why: gemini-3.5-flash-lite appears in the Google Gemini API as a lightweight model. The existing gemini-flash-lite alias pattern (google/gemini-*flash*lite*) does cover it via wildcard. However, a versioned alias may be useful for workflows pinning to this specific generation. Low priority — existing gemini-flash-lite alias already covers this via wildcard.

This is optional; the existing wildcard covers it. No change strictly required.

3. Verify sonnet-6x alias completeness

What: Review the sonnet-6x alias against newly confirmed live Anthropic models.

Why: claude-opus-4.5, claude-opus-4.6, claude-opus-4.7, claude-opus-4.8, and claude-opus-5 are all confirmed live in the Copilot reflect data. The opus alias (copilot/*opus*) covers all of these. The sonnet-6x alias covers sonnet 4.5, 4.6, 5 — all confirmed present. No changes needed.

Full Model Lists by Provider

Anthropic (direct API, 10 models):

• claude-haiku-4-5-20251001
• claude-opus-4-1-20250805
• claude-opus-4-5-20251101
• claude-opus-4-6
• claude-opus-4-7
• claude-opus-4-8
• claude-opus-5
• claude-sonnet-4-5-20250929
• claude-sonnet-4-6
• claude-sonnet-5

Copilot reflect (37 models):

• claude-haiku-4.5, claude-opus-4.5, claude-opus-4.6, claude-opus-4.7, claude-opus-4.8, claude-opus-5
• claude-sonnet-4.5, claude-sonnet-4.6, claude-sonnet-5
• gemini-2.5-pro, gemini-3-flash-preview
• gpt-3.5-turbo, gpt-3.5-turbo-0613, gpt-4, gpt-4-0613, gpt-4-o-preview, gpt-4.1, gpt-4.1-2025-04-14
• gpt-4o, gpt-4o-2024-05-13, gpt-4o-2024-08-06, gpt-4o-2024-11-20, gpt-4o-mini, gpt-4o-mini-2024-07-18
• gpt-5-mini, gpt-5.3-codex, gpt-5.4, gpt-5.4-mini, gpt-5.5
• gpt-5.6-luna, gpt-5.6-sol, gpt-5.6-terra
• mai-code-1-flash-picker
• text-embedding-3-small, text-embedding-3-small-inference, text-embedding-ada-002
• trajectory-compaction

Gemini (direct API, 50 models, notable chat models):

• antigravity-preview-05-2026
• gemini-2.0-flash, gemini-2.0-flash-001, gemini-2.0-flash-lite, gemini-2.0-flash-lite-001
• gemini-2.5-flash, gemini-2.5-flash-image, gemini-2.5-flash-lite, gemini-2.5-pro
• gemini-3-flash-preview, gemini-3-pro-image, gemini-3-pro-preview
• gemini-3.1-flash-image, gemini-3.1-flash-lite, gemini-3.1-pro-preview
• gemini-3.5-flash, gemini-3.5-flash-lite
• gemini-3.6-flash
• gemini-omni-flash-preview
• lyria-3-pro-preview
• nano-banana-pro-preview
• (plus embeddings, TTS, robotics, deep-research models)

OpenAI (direct API, 219 models total; many internal/experimental):

• Relevant public models include gpt-4, gpt-4-0613, gpt-3.5-turbo plus a large number of experimental and internal model variants not suitable for alias coverage

Notes

• Alias coverage is comprehensive: All new Claude Opus (4.5–5) and Sonnet (4.5–5) generations are covered by existing copilot/*opus* and copilot/*sonnet* wildcard patterns.
• All GPT-5.x variants (gpt-5.4, gpt-5.4-mini, gpt-5.5, gpt-5.6-luna/sol/terra, gpt-5.3-codex, gpt-5-mini) are covered by existing numbered aliases.
• Billing multipliers unavailable: The Copilot reflect API returned billing.multiplier: null for all 37 models; pricing validation was done against the docs table which covers OpenAI GPT-5.x models only.
• gpt-5.4-nano appears in the Copilot live endpoint and in the pricing table ($0.20/$1.25) but is on the deprecated/excluded list per task instructions — excluded from proposals.
• trajectory-compaction is the only live Copilot model with no alias coverage. Adding an alias is low-risk and useful for agentic workflows that need conversation compaction.

Warning

Firewall blocked 7 domains

The following domains were blocked by the firewall during workflow execution:

• accounts.google.com
• android.clients.google.com
• clients2.google.com
• collector.githubapp.com
• contentautofill.googleapis.com
• safebrowsingohttpgateway.googleapis.com
• www.google.com

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "accounts.google.com"
    - "android.clients.google.com"
    - "clients2.google.com"
    - "collector.githubapp.com"
    - "contentautofill.googleapis.com"
    - "safebrowsingohttpgateway.googleapis.com"
    - "www.google.com"

See Network Configuration for more information.

Generated by 📦 Daily Model Inventory Checker · sonnet46 · 113.6 AIC · ⌖ 9.06 AIC · ⊞ 8.8K · ◷

• expires on Aug 3, 2026, 4:23 PM UTC-08:00