Function calling in Chat Completions
Models can be given function signatures and return schema-matching JSON arguments; the same update cut gpt-3.5-turbo input pricing 25% and embeddings 75%.
gpt-4-0613 and gpt-3.5-turbo-0613 were fine-tuned to decide when to call a developer-described function and emit JSON arguments. Added a 16K-context gpt-3.5-turbo. Became the basis for tool use in OpenAI's API.
- Date
- Tuesday, 13 June 2023
- Lab
- OpenAI
- Kind
- feature
- Access
- closed API
Figures
| Measure | Value | Measured by |
|---|---|---|
| gpt-3.5-turbo input price after cut | $0.0015 per 1K tokens 25% reduction; output $0.002 per 1K | company |
| Ada embedding price after cut | $0.0001 per 1K tokens 75% reduction | company |
Simon Willison's post (title only was seen in his tag index) corroborates the date.
Sources
- voicebot.ai/2023/06/13/openai-upgrades-gpt-4-and-gpt-3-5-turbo-models-reduces-api-prices/
- simonwillison.net/2023/Jun/13/function-calling/
This record was checked against its sources on 6 October 2026. How we check