chore: update model catalog from bot issues

[github-actions]

Automated daily batch of model catalog updates from bot issues.

Included issues

• Closes [BOT ISSUE] OpenAI: add missing pricing for chat-latest #835: [BOT ISSUE] OpenAI: add missing pricing for chat-latest

Summary

Issue	Provider	Primary model	Changed models	Added models	Updated models	Verification sources
#835	openai	chat-latest	chat-latest	None	chat-latest	1

Verified metadata

#835: [BOT ISSUE] OpenAI: add missing pricing for chat-latest

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
chat-latest	GPT Chat Latest		openai, azure	openai	chat	input=128000, output=16384	in/out=5/30 per 1M; cache read=0.5 per 1M	multimodal=true; reasoning=true

Verification notes

Verification

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	Jun 20, 2026 11:03am

[github-actions]

Codex (@codex) review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 39119e705a

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]

Avoid routing OpenAI-format Bedrock models through this proxy

These catalog entries are format: "openai", so fetchModel dispatches them to fetchOpenAI, and packages/proxy/src/proxy.ts currently rejects secret.type === "bedrock" with Bedrock does not support OpenAI format. Advertising only bedrock here means credential lookup/fallback via getModelEndpointTypes will select a Bedrock secret for openai.gpt-5.5, openai.gpt-5.4, or xai.grok-4.3, then fail before making the upstream request; either Bedrock OpenAI-format support needs to be implemented in this proxy or these endpoint mappings should not be exposed here.

Useful? React with 👍 / 👎.

[github-actions]

Codex Review Response

Issue addressed

Codex flagged that openai.gpt-5.5, openai.gpt-5.4, and xai.grok-4.3 in model_list.json had "format": "openai" but only listed "bedrock" as their available provider. The proxy rejects Bedrock secrets in fetchOpenAI with "Bedrock does not support OpenAI format", so these models would fail at request time.

Fix applied

Changed "format": "openai" to "format": "converse" for all three models in packages/proxy/schema/model_list.json. This matches the convention used by every other Bedrock-only model in the catalog (e.g., openai.gpt-oss-120b-1:0, openai.gpt-oss-safeguard-120b, moonshot.kimi-k2-thinking).

No changes were needed in packages/proxy/schema/index.ts — the endpoint type mappings (["bedrock"]) are correct.