# o1 vs o3

> o3 is the stronger model overall, scoring 47.4 to 40.6 on the Noometry Index.

- Canonical page: https://noometry.com/compare/o1-vs-o3
- Last updated: 2026-10-10
- Shared benchmarks: 41

## Summary

- They share 41 benchmarks with published results for both. o1 scores higher in 1 category and o3 in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where o3 leads 48.9 to 34.9.
- The biggest single-benchmark swing is ARC-AGI-1: 30.7% for o1 and 60.8% for o3.
- o3 is cheaper at $2 / $8 per million input/output tokens, against $15 / $60 for o1.

## Snapshot

| | o1 | o3 |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 40.6 | 47.4 |
| Rank | 140 | 53 |
| Context | 200K | 200K |
| Input $/M | $15 | $2 |
| Output $/M | $60 | $8 |
| Weights | Proprietary | Proprietary |

## Coding

- o1: 45.7 (#64)
- o3: 47.3 (#57)

| Benchmark | o1 | o3 |
|---|---|---|
| Aider Polyglot | 61.7% | 81.3% |
| WeirdML | 47.6% | 52.4% |
| LMArena Coding | 1367 | 1408 |
| CadEval | 56% | 74% |
| SWE-bench Verified | — | 62.3% |
| SWE-bench Verified (bash only) | — | 58.4% |
| GSO | — | 8.8% |
| LiveBench Coding | 69.7% | — |
| ALE-Bench | — | 933.55 |
| HumanEval+ | 89% | — |
| MBPP+ | 80.2% | — |

## Agentic & Tool Use

- o1: 24.3 (#111)
- o3: 34.3 (#42)

| Benchmark | o1 | o3 |
|---|---|---|
| METR Time Horizons | 51.1% | 65.4% |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| Cybench | 10% | — |
| DeepResearch Bench | — | 45.2% |
| OSWorld | — | 23% |
| LMArena Search | — | 1144 |

## Reasoning

- o1: 27.5 (#103)
- o3: 32.1 (#66)

| Benchmark | o1 | o3 |
|---|---|---|
| SimpleBench | 41.7% | 53.1% |
| ARC-AGI-1 | 30.7% | 60.8% |
| Chess Puzzles | 15% | 38% |
| EnigmaEval | 5.7% | 13.1% |
| LMArena Hard Prompts | 1372 | 1402 |
| DTBench | 74.7% | 84.8% |
| LMCA | 22.3% | 39.7% |
| Epoch Capabilities Index | 141.92 | 146.88 |
| ARC-AGI-2 | — | 6.5% |
| Kagi LLM Benchmark | — | 67.6% |
| CritPt | — | 1.4% |
| LiveBench Reasoning | 91.6% | — |
| Mystery Game Puzzles | — | 29% |
| LiveBench Data Analysis | 65.5% | — |
| ForecastBench | — | 62.5 |
| LiveBench | 75.7% | — |

## Math

- o1: 34.9 (#150)
- o3: 48.9 (#49)

| Benchmark | o1 | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 14.7% | 33.3% |
| OTIS Mock AIME 2024-2025 | 73.3% | 84.4% |
| LMArena Math | 1388 | 1426 |
| MATH Level 5 | 94.7% | 97.8% |
| FrontierMath (Feb 2025 set) | 9.3% | 18.7% |
| Omni-MATH | — | 71.4% |
| LiveBench Math | 80.3% | — |
| FrontierMath Tier 4 (v1) | — | 2.1% |

## Knowledge

- o1: 41.5 (#101)
- o3: 54.5 (#47)

| Benchmark | o1 | o3 |
|---|---|---|
| GPQA Diamond | 76.8% | 81.8% |
| Humanity's Last Exam | 8% | 20.3% |
| SimpleQA Verified | 41.1% | 49.4% |
| Confabulations | 11.7% | 14.4% |
| LMArena Expert | 1362 | 1403 |
| MMLU-Pro | — | 85.9% |
| GPQA (HELM) | — | 75.3% |

## Multimodal

- o1: 33.6 (#87)
- o3: 40.7 (#35)

| Benchmark | o1 | o3 |
|---|---|---|
| LMArena Vision | 1168 | 1214 |
| GeoBench | 80% | 74% |
| VPCT | 37% | 52% |
| SpatialViz-Bench | 41.4% | — |

## Multilingual

- o1: 48.8 (#137)
- o3: 51.9 (#99)

| Benchmark | o1 | o3 |
|---|---|---|
| LMArena Non-English | 1359 | 1402 |
| LMArena Chinese | 1394 | 1438 |
| LMArena French | 1343 | 1430 |
| LMArena German | 1337 | 1420 |
| LMArena Japanese | 1346 | 1404 |
| LMArena Korean | 1396 | 1371 |
| LMArena Russian | 1356 | 1407 |
| LMArena Spanish | 1345 | 1395 |

## Instruction Following

- o1: 75.0 (#82)
- o3: 73.0 (#122)

| Benchmark | o1 | o3 |
|---|---|---|
| LMArena Instruction Following | 1368 | 1368 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | — | 86.9% |

## Long Context

- o1: 50.3 (#9)
- o3: 53.2 (#6)

| Benchmark | o1 | o3 |
|---|---|---|
| Fiction.LiveBench | 83.3% | 88.9% |
| LMArena Longer Query | 1378 | 1372 |
| CL-bench | — | 17.8% |

## Writing & Preference

- o1: 55.8 (#139)
- o3: 63.7 (#62)

| Benchmark | o1 | o3 |
|---|---|---|
| LMArena Text | 1366 | 1410 |
| LMArena Creative Writing | 1347 | 1359 |
| Short-Story Creative Writing | 70.2% | 83.9% |
| LMArena Multi-Turn | 1368 | 1405 |
| EQ-Bench Creative Writing | — | 1676 |
| WildBench | — | 86.1% |
| LiveBench Language | 65.4% | — |

## FAQ

### Is o1 better than o3?

o3 is the stronger model overall, scoring 47.4 to 40.6 on the Noometry Index.

### Which is cheaper, o1 or o3?

o3 is cheaper. It lists at $2 per million input tokens and $8 per million output tokens; o1 lists at $15 and $60.

### Is o1 or o3 better for coding?

o3 scores higher on coding benchmarks: 47.3 versus 45.7 in the Noometry coding category.

### Which has the bigger context window?

Both accept 200K tokens.

### How many benchmarks do o1 and o3 share?

41 benchmarks have published results for both models. o1 has 52 scored results on Noometry and o3 has 63.