using openai to call openai.Completion.create the response is shorter than what chatgpt returns

Question

I have a paid account for chatgpt and using the latest 4 version, when I ask it to answer a question it send back a response that is like 4k characters. When I use the api for same question, I get back a much shorter response that is about 1k characters. I decided to ask chatgpt what parameters to use in the api call to get back the same response and I pasted it below. It used max tokens as 30 so I upped it to 4k, but the response is still too short. I then modified the engine to use davinci003 from 002 with no luck. Any ideas on what I can change, I spent about 2 hours researching and still can't get it to work.

def generate_completion(prompt):
    response = openai.Completion.create(
      engine="text-davinci-003",
      prompt=prompt,
      max_tokens=4000,
      n=1,
      stop=None,
      temperature=0.8,
      top_p=1,
      frequency_penalty=0,
      presence_penalty=0,
    )
    generated_text = response.choices[0].text.strip()
    return generated_text

Answer 1

You need to use Chat instead of Completion.

For Python, the example code is as follows:

completion = openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
    {"role": "user", "content": "Hello!"}
  ]
)

If you have access to GPT-4 API, then you can change the model to gpt-4. Note that having a paid account doesn't guarantee access to GPT-4 API. Otherwise, you can use model gpt-3.5-turbo or gpt-3.5-turbo-0301, which is accessible to all but less powerful than the gpt-4 model.