The chat completions API has always allowed clients to pass a null value in the max_tokens parameter. For example, this request yields a valid chat completion: curl "https://api.openai.com/v1/chat/completions" \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $OPENAI_API_KEY…

Looks like they’ve updated the documentation now and explain the new parameter max_completion_tokens. See: Controlling costs section of the reasoning models API reference, where max_tokens has been deprecated

Inconsistent handling of `max_tokens` parameter in chat completions API

API Bugs

kbravh October 23, 2024, 8:08pm 2

Looks like they’ve updated the documentation now and explain the new parameter max_completion_tokens. See:

Controlling costs section of the reasoning models
API reference, where max_tokens has been deprecated

1 Like

Topic		Replies	Views
Annoying error if max_completion_tokens is "too high" Feedback chat-completion , token	1	191	February 3, 2025
Documented max_token default is incorrect for gpt-4-vision-preview API gpt-4	2	2555	November 22, 2023
Chat completions API - max_tokens default value is missing API api , chat-completion	1	3069	July 3, 2024
Max tokens chat completion gpt4o API gpt-4o	4	17092	September 5, 2024
Clarification about max_completion_tokens rate-limiting API rate-limit , o1-preview	4	747	October 10, 2024

Inconsistent handling of `max_tokens` parameter in chat completions API

Related topics