LLMAPI only
o3-mini
OpenAI
Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error
Superseded by o4-mini
Specifications
- Type
- LLM
- Access
- API only
- Released
- January 31, 2025
- License
- Proprietary
- Context window
- 200,000 tokens
- Knowledge cutoff
- September 30, 2023
- Output speed
- 115 tok/s
- Latency (TTFT)
- 5.2s
- Input
- Text
- Output
- Text
- API pricing
- $1.10 in · $4.40 out / 1M tokens
Capabilities
Function callingStructured outputWeb search
Benchmarks
Reasoning
| GPQA Diamond | 77.2 |
Coding
| Aider Polyglot | 66.7 |
| SWE-bench Verified | 49.3 |
| Aider Polyglot Edit | 60.4 |
| LiveCodeBench | 73.4 |
Math
| AIME 2024 | 87.3 |
| FrontierMath | 9.2 |
| MATH | 97.9 |
| MGSM | 92 |
| MATH-500 | 98.5 |
Agents
Our take
A small, fast reasoning model with adjustable effort levels — strong on STEM and coding at a fraction of o1’s cost. It made inference-time reasoning cheap enough for everyday use.