Inconsistent handling of `max_tokens` parameter in chat completions API

API Bugs
1

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" \
    -d '{
        "model": "gpt-4o-mini",
        "messages": [
            {
                "role": "system",
                "content": "You are a helpful assistant."
            },
            {
                "role": "user",
                "content": "Write a haiku that explains the concept of recursion."
            }
        ],
        "max_tokens": null
    }'
{
  "id": "chatcmpl-ADAV8Ndcsz7ebzuD2DnJ59RY9RffC",
  "object": "chat.completion",
  "created": 1727701994,
  "model": "gpt-4o-mini-2024-07-18",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Loops within oneself,  \nA path that leads back again—  \nEchoes of the truth.  ",
        "refusal": null
      },
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 28,
    "completion_tokens": 20,
    "total_tokens": 48,
    "completion_tokens_details": {
      "reasoning_tokens": 0
    }
  },
  "system_fingerprint": "fp_f85bea6784"
}

However, if you try to do this with the new o1 models, the request fails with a 400 error:

curl "https://api.openai.com/v1/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OPENAI_API_KEY" \
    -d '{
        "model": "o1-mini",
        "messages": [
            {
                "role": "system",
                "content": "You are a helpful assistant."
            },
            {
                "role": "user",
                "content": "Write a haiku that explains the concept of recursion."
            }
        ],
        "max_tokens": null
    }'
{
  "error": {
    "message": "Invalid type for 'max_tokens': expected an unsupported value, but got null instead.",
    "type": "invalid_request_error",
    "param": "max_tokens",
    "code": "invalid_type"
  }
}

According to the documentation for the reasoning models, " The max_tokens parameter continues to function as before for all previous models."

But the behavior outlined above is inconsistent between the previous models and the new o1 models, and forces clients to conditionally pass different values in this parameter depending on what model they’re using.