Repository navigation
Support OpenAI reasoning models via OpenAI API #792
Description
Is your feature request related to a problem? Please describe.
chatGPTBox currently can't use OpenAI's o1 series reasoning models due to some API spec differences. One key issue is that the max_tokens parameter has been deprecated and replaced by max_completion_tokens, which leads to incompatibility with the new models.
Describe the solution you'd like
To support the o1 series models, the following changes and considerations should be made based on the OpenAI API and beta limitations:
- Replace
max_tokenswithmax_completion_tokens. - Limitations during beta:
- Only
userandassistantmessage types are allowed (nosystemmessages). - Streaming, tools, and function calling are not supported.
- Parameters such as
temperature,top_p, andnare fixed at 1. presence_penaltyandfrequency_penaltyare fixed at 0.
- Only
Reacted by Copilot
Activity
Mohamed3nan commentedon Oct 1, 2024
It's working if you just add the name of the model :D
[-]Support OpenAI o1 series models[/-][+]Support OpenAI o1 series models via OpenAI API[/+]PeterDaveHello commentedon Oct 1, 2024
Hey @Mohamed3nan,
Thanks for pointing that out! I should probably update the title to be more precise. This issue is specifically about accessing the o1 series models via the API. When using the web, these issues don't occur because the access method is different.
[-]Support OpenAI o1 series models via OpenAI API[/-][+]Support OpenAI reasoning models via OpenAI API[/+]PeterDaveHello commentedon Aug 23, 2025
At least o3, o3-mini, o4-mini, gpt-5, gpt-5-mini, and gpt-5-nano should also be supported and tested.