Closed indigoviolet closed 1 year ago
Looks like it's working.
Try both and see the difference:
ChatGPT(3.5-turbo)> What version of GPT are you?
As an AI language model, I don't run on a specific version of GPT. I am developed by OpenAI and use their latest GPT models for natural language generation and processing. However, I am not designed to provide technical information about the specific version of GPT running behind the scenes, but rather to provide contextual and informative responses to your questions.
ChatGPT(4)> What version of GPT are you?
I am based on OpenAI's GPT-3 model, which is the third iteration of the Generative Pre-trained Transformer (GPT) architecture. The GPT-3 model is known for its large size and impressive capabilities in understanding context and generating human-like text. Keep in mind that my responses are generated through a platform that interacts with GPT-3, but the specific version details might not be disclosed.
Another way to confirm:
(setq shell-maker-logging t) and view the curl command from *chatgpt-log*
buffer.
curl https://api.openai.com/v1/chat/completions --fail-with-body --no-progress-meter -m 180 -H Content-Type: application/json -H Authorization: Bearer SK-REDACTED-OPENAI-KEY -d {"stream":true,"model":"gpt-4","messages":[{"role":"user","content":"What version of GPT are you?"}]}
Before I was allowlisted for gpt-4
those requests would flat out fail with error. If I remember correctly, with an unrecognized model.
OK, I do see that the curl has "model":"gpt-4-0314"
, I guess the output can't be trusted :)
I used
chatgpt-shell-swap-model-version
: