Repository navigation
Codex provider cannot use gpt-5.6-luna model #2643
Description
Bug Description
I'm getting this error when trying to evaluate the new GPT 5.6 Luna model:
2026-07-10 04:59:31,335 - ERROR - hindsight_api.engine.providers.codex_llm - Codex HTTP error after 3 attempts: Status 404, Detail: {
"error": {
"message": "Model not found gpt-5.6-luna",
"type": "invalid_request_error",
"param": "model",
"code": null
}
}
2026-07-10 04:59:31,335 - WARNING - hindsight_api.engine.memory_engine - LLM connection verification failed for 'retain' config: Codex LLM connection verification failed for gpt-5.6-luna: Client error '404 Not Found' for url 'https://chatgpt.com/backend-api/codex/responses'
I can confirm the model is usable in Codex CLI with the same account. Sol and Terra both work, only Luna is throwing this error.
Steps to Reproduce
Use gpt-5.6-luna model for any task, like HINDSIGHT_API_RETAIN_LLM_MODEL=gpt-5.6-luna
Expected Behavior
The model should be usable
Actual Behavior
See the error log above
Version
0.8.3
LLM Provider
OpenAI
Reacted by JiehoonKwak
Activity
JiehoonKwak commentedon Jul 10, 2026
Reproduced on current main and opened #2647 .
The 404 was caused by stale client identity headers, not the model slug. Added the first-party Codex originator and shaped User-Agent through a helper, so it leaves request bodies and the default model unchanged.