o3-mini brings the o3 series' advanced reasoning to a smaller, faster, cheaper package. Released in January 2025, it achieves state-of-the-art results on competitive coding (Codeforces) and mathematics benchmarks at a fraction of the cost of the full o3 model. It supports configurable reasoning effort (low/medium/high) to trade off latency against quality.
Key capabilities
- Configurable reasoning effort — low (fast), medium, or high (most accurate)
- SOTA on AIME math competition — surpasses o1 on several benchmarks
- Tool/function calling — suitable for multi-step agentic tasks
- 200K context window — same as o1
What it's best for
Developers who need o1-class reasoning quality for coding and STEM tasks but want faster responses and lower per-token costs. Particularly strong for competitive programming, mathematical proofs, and structured data extraction.