Skip to content

feat: add GPT-5.4 Codex fast mode models (spark variants)#12114

Draft
roomote-v0[bot] wants to merge 1 commit intomainfrom
feature/gpt-5.4-codex-fast-mode
Draft

feat: add GPT-5.4 Codex fast mode models (spark variants)#12114
roomote-v0[bot] wants to merge 1 commit intomainfrom
feature/gpt-5.4-codex-fast-mode

Conversation

@roomote-v0
Copy link
Copy Markdown
Contributor

@roomote-v0 roomote-v0 bot commented Apr 14, 2026

This PR attempts to address Issue #12112.

Summary

Adds three new model entries to the OpenAI Codex provider for GPT-5.4 fast mode support:

  • gpt-5.4-codex: Full GPT-5.4 optimized for agentic coding (1M+ context window, 128k max output)
  • gpt-5.4-codex-spark: Fast, text-only GPT-5.4 variant (128k context, 8k output) - the "fast mode"
  • gpt-5.4-mini-codex-spark: Fast, text-only GPT-5.4 Mini variant (128k context, 8k output)

These follow the same pattern as the existing gpt-5.3-codex and gpt-5.3-codex-spark models, with spark variants providing faster responses by trading off context window size and image support.

Changes

  • packages/types/src/providers/openai-codex.ts: Added 3 new model definitions
  • src/api/providers/__tests__/openai-codex.spec.ts: Added tests for the new models

Test plan

  • All 15 tests pass in openai-codex.spec.ts
  • Lint and type checks pass across the monorepo

Feedback and guidance are welcome.

Interactively review PR in Roo Code Cloud

Adds three new model entries to the OpenAI Codex provider:
- gpt-5.4-codex: Full GPT-5.4 optimized for agentic coding (1M context)
- gpt-5.4-codex-spark: Fast, text-only GPT-5.4 variant (128k context, 8k output)
- gpt-5.4-mini-codex-spark: Fast, text-only GPT-5.4 Mini variant (128k context, 8k output)

Follows the same pattern as existing gpt-5.3-codex and gpt-5.3-codex-spark models.

Closes #12112
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant