chore: update model catalog from bot issues#904
Conversation
github-actions
Bot
requested review from
Alex Z (CLowbrow),
aswink,
Caitlin Pinn (cpinn),
Erin McNulty (erin2722) and
Ken Jiang (knjiang)
|
Codex (@codex) review |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1f70a30da5
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".
| ], | ||
| "max_input_tokens": 1000000, | ||
| "max_output_tokens": 64000, | ||
| "max_output_tokens": 128000, |
There was a problem hiding this comment.
When callers use us.anthropic.claude-sonnet-4-6 or global.anthropic.claude-sonnet-4-6 directly, their own model entries still say max_output_tokens: 64000 even though this top-level model now lists them as fallback aliases and is raised to 128k. The Anthropic adapter defaults max_tokens from the selected model spec (openaiParamsToAnthropicMesssageParams), so requests that omit max_tokens for those Bedrock aliases will still be capped at 64k; update the two alias entries along with this value.
Useful? React with 👍 / 👎.
| "input_cost_per_mil_tokens": 1.5, | ||
| "output_cost_per_mil_tokens": 7.5, | ||
| "displayName": "Mistral Medium 3.5 (2026-04)", | ||
| "parent": "mistral-medium-3.5", |
There was a problem hiding this comment.
This newly added dated snapshot is parented to mistral-medium-3.5, but the catalog does not inherit reasoning or reasoning_budget from parent entries—getAvailableModels() returns each parsed spec as-is. Because the same commit marks the parent/latest Medium 3.5 ids as reasoning-capable, and Mistral documents Medium 3.5 as supporting reasoning_effort (https://docs.mistral.ai/studio-api/conversations/reasoning), selecting mistral-medium-3-5-26-04 directly will omit that capability; add the same flags to this entry.
Useful? React with 👍 / 👎.
…8k aliases Review fixes for the daily batch: - Remove mistral-medium-3-5-26-04 (#890). The id is not a real Mistral model — both the gateway and api.mistral.ai return 400 "Invalid model: mistral-medium-3-5-26-04" / invalid_model. Mistral serves mistral-medium-3-5 and mistral-medium-3.5, not a -26-04 dated snapshot. Removing the entry also resolves the Codex P2 about the snapshot missing reasoning flags. - Raise us.anthropic.claude-sonnet-4-6 and global.anthropic.claude-sonnet-4-6 max_output_tokens 64000 -> 128000 (Codex P2). #884 raised the other Sonnet 4.6 entries to 128k (correct per Anthropic's model card) but left these regional bedrock aliases at 64000; the Anthropic adapter defaults max_tokens from the selected spec, so callers using the aliases directly would have been capped low. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
1 participant
Automated daily batch of model catalog updates from bot issues.
Included issues
Summary
gemma-4-31bgemma-4-31b2
claude-sonnet-4-6anthropic.claude-sonnet-4-6publishers/anthropic/models/claude-sonnet-4-6claude-sonnet-4-6anthropic.claude-sonnet-4-6publishers/anthropic/models/claude-sonnet-4-62
pixtral-large-2411mistral-large-2411pixtral-large-2411mistral-large-2411mistral-medium-3-5-26-04mistral-medium-3-5-26-042
3
4
mistral-medium-latestmistral-medium-3.5mistral-small-latestmistral-medium-latestmistral-medium-3.5mistral-small-latest2
gpt-5.4-minigpt-5.4-mini-2026-03-17gpt-5.4-nanogpt-5.4-minigpt-5.4-mini-2026-03-17gpt-5.4-nano2
3
gemini-3.5-flashgemini-3.5-flash2
qwen/qwen3-32bmeta-llama/llama-4-scout-17b-16e-instructllama-3.3-70b-versatileqwen/qwen3-32bmeta-llama/llama-4-scout-17b-16e-instructllama-3.3-70b-versatile2
openai/gpt-oss-120bopenai/gpt-oss-120b2
3
Qwen/Qwen3.5-397B-A17BQwen/Qwen3.5-397B-A17Baccounts/fireworks/models/glm-5p2accounts/fireworks/models/glm-5p22
Verified metadata
#883: [BOT ISSUE] Cerebras: add missing gemma-4-31b model
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#884: [BOT ISSUE] Anthropic: update Claude Sonnet 4.6 max_output_tokens from 64000 to 128000
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#885: [BOT ISSUE] Mistral: add deprecation markers for retired pixtral-large-2411 and mistral-large-2411
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#890: [BOT ISSUE] Mistral: add missing mistral-medium-3-5-26-04 snapshot
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#891: [BOT ISSUE] Mistral: add missing reasoning flags for Medium 3.5 and Small 4
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#894: [BOT ISSUE] OpenAI: update gpt-5.4-mini and gpt-5.4-nano max_input_tokens from 272000 to 400000
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#895: [BOT ISSUE] Google: add missing reasoning_budget flag for gemini-3.5-flash
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#896: [BOT ISSUE] Groq: add deprecation markers for 3 models scheduled for shutdown
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
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.
#901: [BOT ISSUE] Baseten: update openai/gpt-oss-120b pricing to match Baseten official rates
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.
#902: [BOT ISSUE] Together: fix Qwen3.5-397B-A17B cached input pricing from $0.06 to $0.35
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.
#903: [BOT ISSUE] Fireworks: fix GLM 5.2 cached input pricing from $0.26 to $0.14
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.