fix: add OpenAI models chat-latest

[github-actions]

fix: add OpenAI models chat-latest

Closes #586

Source issue: #586

Summary

Field	Value
Provider	openai
Primary model	chat-latest
Changed models	chat-latest
Added models	chat-latest
Updated models	None
Verification sources	1 2

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
chat-latest	GPT Chat Latest		openai, azure	openai	chat	input=128000, output=16384	n/a	multimodal=true; reasoning=true

Verification notes

Verification

Fields verified from official sources

Field	Value	Source
format	openai	Inferred from OpenAI provider (matches all existing OpenAI chat models in catalog)
flavor	chat	Azure docs: "Chat Completions API" in capabilities
max_input_tokens	128000	Azure models page: "Context Window: 128,000" — matches existing gpt-5.x-chat-latest catalog entries which all use 128000
max_output_tokens	16384	Azure models page: "Max Output Tokens: 16,384" — matches existing gpt-5.x-chat-latest catalog entries
reasoning	true	Azure models page: "Reasoning" listed in capabilities
multimodal	true	Follows pattern of all existing gpt-5.x-chat-latest entries in catalog; Azure note identifies this as "GPT-5.5 Instant" and gpt-5.5 supports "Text and image processing"
displayName	"GPT Chat Latest"	Follows catalog pattern: gpt-5-chat-latest → "GPT-5 chat", gpt-5.3-chat-latest → "GPT-5.3 chat"; Azure page heading is "GPT-chat-latest"
available_providers	openai, azure	Azure docs confirm gpt-chat-latest deployment name; OpenAI API name confirmed as chat-latest

Fields omitted (not verifiable)

Field	Reason
input_cost_per_mil_tokens	OpenAI pricing page returned 403; Azure pricing page loads dynamically. The original issue claimed $5/1M (matching gpt-5.5) but chat-class models have historically been significantly cheaper ($1.25–$1.75/1M). Cannot verify.
output_cost_per_mil_tokens	Same as above. Original issue claimed $30/1M but unverifiable.
input_cache_read_cost_per_mil_tokens	Same as above. Original issue claimed $0.50/1M but unverifiable.
parent	No parent — this is a rolling alias, not a dated snapshot. Existing gpt-5.x-chat-latest entries also have no parent.

Corrections from original issue body

1. Token limits: Original issue stated max_input_tokens: 1050000 and max_output_tokens: 128000 (gpt-5.5 specs). Azure docs show 128K context / 16K output, matching the chat-class series, not gpt-5.5. Corrected to max_input_tokens: 128000, max_output_tokens: 16384.
2. Pricing: Original issue stated pricing matching gpt-5.5. Removed because (a) cannot verify from any accessible official source and (b) chat-class models historically have different pricing than their parent model families.

sync_models (LiteLLM) cross-check

Field	Proposed value	sync_models value	Note
(all fields)	(as specified)	Model not found in sync_models	chat-latest and gpt-chat-latest are absent from the LiteLLM model_prices_and_context_window_backup.json catalog. No cross-check possible.

sync_models vs proposed update

sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.

Model	Field	Proposed update	sync_models	sync_models source models
chat-latest	catalog entry	present	missing	None

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ai-proxy	Ready	Preview, Comment	May 21, 2026 1:19pm

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 81d2afb8ac

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "Codex (@codex) address that feedback".

[chatgpt-codex-connector]

Map Azure deployment alias for chat-latest

Do not advertise chat-latest as Azure-ready without an Azure alias, because Azure requests without an explicit metadata.deployment derive the deployment path from the model name (packages/proxy/src/proxy.ts uses bodyData.model for /openai/deployments/{...}), which will send chat-latest instead of Azure’s gpt-chat-latest. In that common configuration, calls routed to Azure will fail at runtime (deployment not found) even though this catalog entry marks the model as available on Azure.

Useful? React with 👍 / 👎.