Overview
Davinci-002 is a GPT base model from OpenAI that understands and generates natural language or code. It is not trained with instruction following, so it extends the text in the prompt instead of carrying out instructions, and it uses the legacy text-completion interface rather than chat. OpenAI lists it as deprecated.
Davinci-002 can generate up to 16,384 output tokens per request. It accepts text input and returns text output. Its knowledge cutoff is September 1, 2021.
Capabilities
Davinci-002 reads text and writes text. Because it is a base model, a good prompt looks like the start of the document or code file you want, and the model carries it forward.
Its output limit is 16,384 tokens per request, and its knowledge cutoff is September 1, 2021.
Pricing
Davinci-002 is billed per token: $2.00 per 1M input tokens and $2.00 per 1M output tokens. Davinci-002 is available through RealRelay's OpenAI-compatible API.
Use cases
Use Davinci-002 for completion-style generation: extending an article from its first paragraph, producing text in the style of the examples you place in the prompt, or continuing a half-written function.
It fits existing workloads built on text completion. Chat assistants and tasks that rely on following written instructions belong with models designed for them.
What is Davinci-002?
Davinci-002 is a GPT base model from OpenAI that understands and generates natural language or code. It is not trained with instruction following, so it extends the text in the prompt instead of carrying out instructions, and it uses the legacy text-completion interface rather than chat. OpenAI lists it as deprecated.
How much does Davinci-002 cost?
Davinci-002 costs $2.00 per 1M input tokens and $2.00 per 1M output tokens.
What is the maximum output of Davinci-002?
Davinci-002 can generate up to 16,384 output tokens per request.
What inputs and outputs does Davinci-002 support?
Davinci-002 accepts text input and returns text output.
What is the knowledge cutoff of Davinci-002?
The knowledge cutoff of Davinci-002 is September 1, 2021.
How do I call Davinci-002 through the API?
To call Davinci-002, send a request to RealRelay's OpenAI-compatible API with the model ID openai/davinci-002.