#"gpt-3.5-turbo-0301" being used in completion requests when passing "gpt-3.5-turbo" as the model.

4 messages · Page 1 of 1 (latest)

glossy yacht
#

Is anyone else seeing the API return completions that seem to be using "gpt-3.5-turbo-0301" when they've passed in "gpt-3.5-turbo" . These completions also seem to be truncated. It was my understanding that "gpt-3.5-turbo-0301" has been deprecated.

Here is an example of my completion response:

completion: {
  "choices": [
    {
      "finish_reason": "length",
      "index": 0,
      "message": {
        "content": "When it comes to varnishing outdoor furniture, there are several options to consider. One of the most popular options is spar varnish, which is specifically designed to withstand the harsh environmental conditions that furniture may be exposed to outdoors. Other options include water-based polyurethane and oil-based polyurethane. Here are the pros and cons of each:\n\n1. Spar Varnish:\nPros:\n- Highly resilient and durable\n- High gloss and flexibility\n- Designed to withstand salt spray, intense sun",
        "role": "assistant"
      }
    }
  ],
  "created": 1686499125,
  "id": "chatcmpl-7QHkjPIqwbOL80fBF1D7fTNK2ArGE",
  "model": "gpt-3.5-turbo-0301",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 100,
    "prompt_tokens": 3997,
    "total_tokens": 4097
  }
}
#

Just to add a bit of context, I'm using the OpenAI Python lib, v0.27.8

stark dagger
#

From my understanding, "gpt-3.5-turbo" is the general model name, while "gpt-3.5-turbo-0301" is the latest checkpoint. It's not deprecated.
As for completions being truncated, it seems like you reached the token limit..?🤔
Your prompt was 3997 tokens. That's almost the entire context window of gpt 3.5

glossy yacht
#

Gotcha, thanks for the help.