[Bug] max_token is too large

[Algorithm5838]

Describe the bug
When I used the new gpt-3.5-turbo-1106 model with the updated default max_token value of 8192, I just sent "Hello", I encountered the following error:

{
  "error": {
    "message": "max_tokens is too large: 8192. This model supports at most 4096 completion tokens, whereas you provided 8192.",
    "type": "invalid_request_error",
    "param": "max_tokens",
    "code": null
  }
}

Which is strange because the new model should support 16k. The same thing happens with gpt-4-1106-preview. However, when I use the older version gpt-3.5-turbo-16k, it works flawlessly without any errors. But when I use gpt-4, I just sent "Hello", I get this error:

{
  "error": {
    "message": "This model's maximum context length is 8192 tokens. However, you requested 8256 tokens (64 in the messages, 8192 in the completion). Please reduce the length of the messages or completion.",
    "type": "invalid_request_error",
    "param": "messages",
    "code": "context_length_exceeded"
  }
}

Deployment

• Docker
• Vercel
• Server

[cutejun]

+1

[xiangmingwu]

+1

[pmao-guess]

same problem ， any one can help

[Yidadaa]

fixed

[Algorithm5838]

Please check the discussion #3208.