OpenAIReasoning modelo3-pro
Best configuration ranks #200 of 610 on the BenchLeader Index at 54.4 ±7.7 (high reasoning effort). Last measured 26 Jun 2025stale: no new result in six months. Released 10 Jun 2025.
- Blended price
- $35.00/M
- $20.00 in · $80.00 out
- Output speed
- –
- First answer
- –
- Context
- 200k
- Overall index54
- Reasoning41
- Coding56
- Knowledge62
- Instruction following59
- Multimodal66
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| low | – | – | – | – | $0.032 | Reasoning 38 |
| medium | – | – | – | – | $0.032 | Long context 72 · Reasoning 40 |
| highbest | 54.4 | #200 | – | – | $0.032 | Coding 56 · Instruction following 59 · Knowledge 62 · Multimodal 66 · Reasoning 41 |
| default | 54.1 | #207 | – | – | $0.032 | Agents & tools 18 · Coding 74 · Composite 57 · Instruction following 59 · Knowledge 62 · Reasoning 66 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| GPQA Diamond (AA)not in index | – | – | – | 84.5%#136 | Artificial Analysis |
| Kagi LLM Benchmark | – | – | – | 72.1%#24 | Kagi LLM Benchmark |
| ARC-AGI-1 | 44.3%#127 | 57.0%#114 | 59.3%#109 | – | ARC Prize |
| ARC-AGI-2 | 2.1%#141 | 1.9%#144 | 4.9%#120 | – | ARC Prize |
Coding
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| WeirdML | – | – | 58.2%#49 | – | WeirdML |
| Aider Polyglot | – | – | – | 84.9%#2 | Aider polyglot leaderboard |
Agents & tools
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| MCP Atlas | – | – | – | 44.5%#29 | Scale AI SEAL |
Knowledge
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| PRBench Finance | – | – | – | 49.1%#10 | Scale AI SEAL |
| PRBench Legal | – | – | – | 49.7%#8 | Scale AI SEAL |
| MultiNRC | – | – | 49.0%#10 | – | Scale AI SEAL |
Instruction following
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| MultiChallenge | – | – | 62.4%#8 | – | Scale AI SEAL |
| TutorBench | – | – | – | 54.6%#5 | Scale AI SEAL |
Multimodal
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| VISTA | – | – | 51.6%#6 | – | Scale AI SEAL |
Long context
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| Fiction.LiveBench 120k | – | 88.9%#5 | – | – | Fiction.live |
Composite
| Benchmark | low | medium | high | default | Source |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | 147.4#53 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | – | – | – | 21.9#154 | Artificial Analysis |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.032 | – | – |
| Summarise a 30-page report | 12,000 / 600 | $0.288 | – | – |
| Code edit | 6,000 / 1,500 | $0.240 | – | – |
| Agentic coding session | 60,000 / 4,000 | $1.52 | – | – |
| Structured extraction | 2,000 / 200 | $0.056 | – | – |
See also
Data as of 9 Sept 2026. Compare these configurations.