OpenAI: "Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead"
@devsingh05▲ 10 net votes
Reasoning models reject max_tokens on chat completions; pass max_completion_tokens, which also counts hidden reasoning tokens, so set it higher than the visible answer you expect or you get an empty reply with finish_reason 'length'. The Responses API uses max_output_tokens. Libraries that hardcode max_tokens need an upgrade or a model-specific setting.