Skip to content

[Bug] Codex 0.153.4 rejects 'minimal'/'none' on gpt-6-astra mapped custom models #3775

Description

@leonclab

Proven native-alias normalization landed; reported gateway case remains open

#3804 (0a4e4bc9bf) normalizes custom reasoning lists only for capability-backed canonical Codex-forward aliases. Arbitrary YYLJ gateway capability and the reported Desktop-version case remain unverified.

Verified against dev 5759d9ea2f1e7281cdc01eb9628f2e0a123fb59c. Original report by @leonclab. The attribution record was added in #3811.

Client or integration

Codex App

Area

Catalog / models

Summary

In the latest Codex Desktop runtime (0.153.4, Darwin arm64), the native binary strictly validates supported reasoning efforts for gpt-6-astra and only permits ['low', 'medium', 'high', 'xhigh', 'max']. Supplying minimal or none causes Codex to immediately reject the request before dispatch with:
Unsupported value: 'minimal' is not supported with the 'gpt-6-astra' model. Supported values are: 'low', 'medium', 'high', 'xhigh', and 'max'.

While OpenCodex logs [opencodex] Removed unsupported reasoning efforts: minimal, none, custom models mapped to Astra (e.g., forward proxy YYLJ/gpt-6-astra with legacy customModels[*].reasoningEfforts containing minimal or none) still have those unsupported efforts preserved in opencodex-catalog.json and turn execution contexts, triggering client-side rejection.

Reproduction

  1. Configure a custom model in ~/.opencodex/config.json pointing to an Astra upstream:
{
  "provider": "YYLJ",
  "modelId": "gpt-6-astra",
  "reasoningEfforts": ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
}
  1. Run ocx sync.
  2. In Codex Desktop (version 0.153.4), start a turn or assign a subtask with reasoning effort minimal.
  3. Observe the immediate 400 error:
{"type":"error","error":{"type":"invalid_request_error","code":"unsupported_value","message":"Unsupported value: 'minimal' is not supported with the 'gpt-6-astra' model. Supported values are: 'low', 'medium', 'high', 'xhigh', and 'max'.","param":"reasoning.effort"},"status":400}

Version

@bitkyc08/opencodex 2.43.0

Operating system

macOS 15.5 (Darwin arm64)

Provider and model

custom / gpt-6-astra

Logs or error output

[opencodex] Removed unsupported reasoning efforts: minimal, none
Unsupported value: 'minimal' is not supported with the 'gpt-6-astra' model. Supported values are: 'low', 'medium', 'high', 'xhigh', and 'max'.

Redacted configuration

{
  "customModels": [
    {
      "provider": "YYLJ",
      "modelId": "gpt-6-astra",
      "reasoningEfforts": ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
    }
  ]
}

Checks

  • I searched existing issues and documentation.
  • I removed secrets, tokens, account details, request credentials, and personal data.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workingcatalogModel catalog, slugs, visibility, routed entries

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions