o3-mini
o3-mini, a small reasoning model with function calling and three effort levels, gives free ChatGPT users a reasoning model for the first time.
First small OpenAI reasoning model with function calling, Structured Outputs and developer messages; low, medium and high reasoning effort; no vision. OpenAI says medium effort matches o1 on math, coding and science, with responses 24% faster than o1-mini. API at $1.10/$4.40 per million tokens.
- Date
- Friday, 31 January 2025
- Lab
- OpenAI
- Kind
- model
- Access
- closed API
- Price
- $1.10 input / $4.40 output per 1M tokens (2025-01-31)
Figures
| Measure | Value | Measured by |
|---|---|---|
| API price (input / output per 1M tokens) | $1.10 / $4.40 less than half of GPT-4o at the time | company |
| Response speed vs o1-mini | 24% faster A/B test, medium effort | company |
| Codeforces Elo (o3-mini high / medium) | 2130 / 2036 from an early system-card PDF later replaced; o1 scored 1891 | company |
Shipped 11 days after DeepSeek-R1 (2025-01-20) and exposed a search-enabled mode in ChatGPT; the reasoning-for-all-users shift is widely read as a response to R1 (inference, not stated by OpenAI).
Sources
- openai.com/index/openai-o3-mini/
- simonwillison.net/2025/Jan/31/o3-mini/
- developers.openai.com/api/docs/changelog
This record was checked against its sources on 6 October 2026. How we check