o1
ActiveReasoning-focused model that thinks before answering.
Overview
The o1 model is trained to spend additional inference-time compute on internal reasoning before producing a response. Strong on math, coding, and science benchmarks.
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| AIME 2024Math | 83.3% accuracy | Self-reported OpenAI o1 system card |
| Aider PolyglotCoding | 61.7% pass@2 | Third-party Papers With Code |
| GPQA DiamondReasoning | 78% accuracy | Self-reported OpenAI o1 system card |
| MATHMath | 94.8% accuracy | Self-reported OpenAI o1 system card |
Integrations & tooling support
- Tool calling
- Supported
- Structured outputs
- Supported
Price vs quality
Overpriced
Mid-tier performance at frontier pricing.
- Quality percentile
- 70.8%
- Effective price
- $48.75/1M
- Pricing breakdown
- $15/1M in
$60/1M out
vs 4 benchmarks
/ 1M tokens (input + 3× output)
Community ratings
No ratings yet. Be the first to rate o1.
Rate o1
Sign in to rate and review.
Comments
Sign in to leave a comment.