OpenAI o1
Last updated: August 10, 2026
Specs
| Vendor | OpenAI |
|---|
| Released | December 5, 2024 |
|---|
| Context window | 200K tokens |
|---|
| Input | $15 / M tokens |
|---|
| Output | $60 / M tokens |
|---|
Benchmarks
| mmlu | 88.7 |
|---|
| gpqa | 78 |
|---|
| humanEval | 89.1 |
|---|
| arenaElo | 1380 |
|---|
| sweBench | 48.9 |
|---|
| sweBenchPro | 36 |
|---|
| liveCodeBench | 72 |
|---|
| aiderPolyglot | 61 |
|---|
| terminalBench | 33 |
|---|
| mmluPro | 80 |
|---|
| hle | 8.8 |
|---|
| aime | 83.3 |
|---|
| math500 | 96.4 |
|---|
| mmmu | 75 |
|---|
Strengths
- First frontier reasoning model, proven in production
- Strong on math, physics and competitive coding
- Handles multi-step scientific workflows
Weaknesses
- Superseded by o3 on most metrics
- Still slow and expensive per call
- No tool use in the original release
Best for
- Existing o1-integrated pipelines
- Benchmark-stable workloads
- Customers with negotiated o1 pricing
Compare with other AI models