chore: sync new models#747
Conversation
|
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: 3b9e453083
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".
| "input_cost_per_mil_tokens": 2, | ||
| "output_cost_per_mil_tokens": 6, |
There was a problem hiding this comment.
For grok-4.20-0309-reasoning and the Grok 4.20 aliases changed in this block, the xAI pricing page currently lists the 1M-context models at $1.25 input / $0.20 cached / $2.50 output per 1M tokens, not $2/$6 (https://docs.x.ai/developers/pricing). With these higher values, any cost display or billing estimate derived from this metadata overstates normal Grok 4.20 usage, including the beta/non-reasoning and multi-agent entries below that received the same edit.
Useful? React with 👍 / 👎.
| "max_input_tokens": 202752, | ||
| "max_output_tokens": 128000, | ||
| "max_input_tokens": 202800, | ||
| "max_output_tokens": 202800, |
There was a problem hiding this comment.
This raises the Fireworks GLM-5.1 advertised output limit to the full 202.8K context window, but Z.AI's model docs distinguish a 200K context length from a 128K maximum output, and Fireworks only advertises a 202k context length for accounts/fireworks/models/glm-5p1 (https://docs.z.ai/guides/llm/glm-5.1, https://fireworks.ai/models/fireworks/glm-5p1). Callers that trust max_output_tokens can now request completions above the provider's real output cap and get provider errors or truncation instead of being limited correctly.
Useful? React with 👍 / 👎.
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
3b9e453 to
53aeb01
Compare
|
Codex (@codex) review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 53aeb01723
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".
| "input_cost_per_mil_tokens": 2, | ||
| "output_cost_per_mil_tokens": 6, |
There was a problem hiding this comment.
For grok-4.20-0309-reasoning and the beta/non-reasoning/multi-agent aliases changed in this sync, the current xAI model docs still list a 1,000,000-token context window with $1.25/M input and $2.50/M output pricing, not $2/$6 and 2,000,000. Leaving these higher prices and expanded limits in model_list.json will overstate costs for usage reporting/billing and allow callers to select token budgets that xAI rejects for these models.
Useful? React with 👍 / 👎.
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
53aeb01 to
28bc809
Compare
|
Codex (@codex) review |
|
Codex Review: Didn't find any major issues. More of your lovely PRs please. ℹ️ About Codex in GitHubCodex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
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". |
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
28bc809 to
e7f0731
Compare
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
e7f0731 to
15a4c7c
Compare
|
Codex (@codex) review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 15a4c7cf2f
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_output_tokens": 128000 | ||
| }, | ||
| "claude-opus-4-7-20260416": { | ||
| "format": "openai", |
There was a problem hiding this comment.
When callers select claude-opus-4-7-20260416 with the listed anthropic provider, this openai format makes fetchModel dispatch to fetchOpenAI instead of the Anthropic translation path in fetchAnthropicChatCompletions (packages/proxy/src/proxy.ts). That sends the OpenAI chat path/body to Anthropic rather than converting it to /messages, so this newly added alias is likely unusable even though the neighboring Claude aliases are marked anthropic.
Useful? React with 👍 / 👎.
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
15a4c7c to
0d673ac
Compare
|
Codex (@codex) review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0d673ace3c
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_output_tokens": 128000 | ||
| }, | ||
| "claude-opus-4-7-20260416": { | ||
| "format": "openai", |
There was a problem hiding this comment.
When clients call the Anthropic-compatible proxy endpoints (/anthropic/messages or /anthropic/v1/messages) with claude-opus-4-7-20260416, this openai format makes fetchModel dispatch to fetchOpenAI instead of fetchAnthropic; with an Anthropic secret that builds URLs like https://api.anthropic.com/v1/anthropic/v1/messages and skips the Anthropic headers/translation path. The parent and other dated Claude Anthropic models in this catalog use format: "anthropic", so this new entry will fail for Anthropic SDK/native-message callers even though it is advertised as available on the anthropic provider.
Useful? React with 👍 / 👎.
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
0d673ac to
ca3fc9b
Compare
|
Codex (@codex) review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ca3fc9b76b
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_output_tokens": 128000 | ||
| }, | ||
| "claude-opus-4-7-20260416": { | ||
| "format": "openai", |
There was a problem hiding this comment.
When callers select claude-opus-4-7-20260416, AvailableEndpointTypes now routes it to an Anthropic secret, but this entry declares format: "openai". fetchModel dispatches solely on modelSpec.format, so these requests go through fetchOpenAI and POST to the Anthropic base URL with the OpenAI /chat/completions path instead of the Anthropic messages translation path. This makes the newly added model unusable with Anthropic keys; the other Claude dated Anthropic entries use format: "anthropic".
Useful? React with 👍 / 👎.
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
ca3fc9b to
3981c14
Compare
|
Codex (@codex) review |
github-actions
Bot
force-pushed
the
chore/sync-new-models
branch
from
3981c14 to
2ae8043
Compare
|
Codex (@codex) review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2ae8043886
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_output_tokens": 128000 | ||
| }, | ||
| "claude-opus-4-7-20260416": { | ||
| "format": "openai", |
There was a problem hiding this comment.
This model is routed to the Anthropic provider via available_providers/AvailableEndpointTypes, but fetchModel dispatches solely on modelSpec.format; with format: "openai", requests for claude-opus-4-7-20260416 go through fetchOpenAI, which builds https://api.anthropic.com/v1/chat/completions with OpenAI-shaped payloads instead of the Anthropic Messages path. The adjacent Claude dated snapshots use format: "anthropic", so this newly advertised model will fail for users selecting it directly.
Useful? React with 👍 / 👎.
1 participant
Automated sync of model metadata and newly available models from the approved provider allowlist.