# Claude Opus 5 vs o3

> Claude Opus 5 is the stronger model overall, scoring 67.7 to 47.4 on the Noometry Index. o3 costs 2.9× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/claude-opus-5-vs-o3
- Last updated: 2026-10-10
- Shared benchmarks: 34

## Summary

- They share 34 benchmarks with published results for both. Claude Opus 5 scores higher in 9 categories and o3 in 1 category; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Claude Opus 5 leads 77.1 to 32.1.
- The biggest single-benchmark swing is ARC-AGI-2: 90.4% for Claude Opus 5 and 6.5% for o3.
- o3 is cheaper at $2 / $8 per million input/output tokens, against $5 / $25 for Claude Opus 5.
- Claude Opus 5 accepts more context: 1M tokens versus 200K.

## Snapshot

| | Claude Opus 5 | o3 |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 67.7 | 47.4 |
| Rank | 4 | 53 |
| Context | 1M | 200K |
| Input $/M | $5 | $2 |
| Output $/M | $25 | $8 |
| Weights | Proprietary | Proprietary |

## Coding

- Claude Opus 5: 67.0 (#5)
- o3: 47.3 (#57)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| WeirdML | 91.8% | 52.4% |
| LMArena Coding | 1533 | 1408 |
| ALE-Bench | 2,165 | 933.55 |
| SWE-bench Verified | — | 62.3% |
| DeepSWE | 73.6% | — |
| FrontierCode | 53.4% | — |
| SWE-bench Verified (bash only) | — | 58.4% |
| Aider Polyglot | — | 81.3% |
| CursorBench | 46.6% | — |
| LMArena WebDev | 1694 | — |
| FrontierSWE | 52% | — |
| SciCode | 56.4% | — |
| GSO | — | 8.8% |
| CadEval | — | 74% |

## Agentic & Tool Use

- Claude Opus 5: 55.5 (#1)
- o3: 34.3 (#42)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| APEX-Agents | 65.8% | — |
| Berkeley Function Calling Leaderboard | — | 63% |
| OSWorld 2.0 | 31.4% | — |
| GDPval | — | 30.8% |
| τ²-bench Banking | 48.7% | — |
| DeepResearch Bench | — | 45.2% |
| OSWorld | — | 23% |
| PostTrainBench | 35% | — |
| BALROG | 63.4% | — |
| GBAEval | 79.6% | — |
| GDP.pdf | 24% | — |
| LMArena Search | — | 1144 |
| METR Time Horizons | — | 65.4% |
| Vending-Bench 2 | 11,182 | — |

## Reasoning

- Claude Opus 5: 77.1 (#3)
- o3: 32.1 (#66)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| ARC-AGI-2 | 90.4% | 6.5% |
| SimpleBench | 80.6% | 53.1% |
| ARC-AGI-1 | 97.5% | 60.8% |
| CritPt | 29.1% | 1.4% |
| Chess Puzzles | 42% | 38% |
| LMArena Hard Prompts | 1525 | 1402 |
| Mystery Game Puzzles | 59% | 29% |
| DTBench | 97.9% | 84.8% |
| LMCA | 64.5% | 39.7% |
| Epoch Capabilities Index | 162.67 | 146.88 |
| Kagi LLM Benchmark | — | 67.6% |
| NYT Connections (extended) | 94.3% | — |
| EnigmaEval | — | 13.1% |
| EBR-Bench | 45.7% | — |
| Bench to the Future 3 | 0.12 | — |
| ForecastBench | — | 62.5 |

## Math

- Claude Opus 5: 86.1 (#6)
- o3: 48.9 (#49)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 85.6% | 33.3% |
| OTIS Mock AIME 2024-2025 | 98.9% | 84.4% |
| LMArena Math | 1533 | 1426 |
| FrontierMath Tier 4 | 73.2% | — |
| ProofBench | 99% | — |
| Omni-MATH | — | 71.4% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 18.7% |
| FrontierMath Tier 4 (v1) | — | 2.1% |

## Knowledge

- Claude Opus 5: 66.8 (#8)
- o3: 54.5 (#47)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| GPQA Diamond | 93.9% | 81.8% |
| SimpleQA Verified | 59.9% | 49.4% |
| LMArena Expert | 1558 | 1403 |
| Humanity's Last Exam | — | 20.3% |
| MMLU-Pro | — | 85.9% |
| Confabulations | — | 14.4% |
| GPQA (HELM) | — | 75.3% |

## Multimodal

- Claude Opus 5: 51.1 (#6)
- o3: 40.7 (#35)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| LMArena Vision | 1321 | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
| Blueprint-Bench 2 | 30.4% | — |
| Furniture Assembly | 60.8% | — |
| LMArena Document | 1516 | — |

## Multilingual

- Claude Opus 5: 58.8 (#3)
- o3: 51.9 (#99)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| LMArena Non-English | 1500 | 1402 |
| LMArena Chinese | 1575 | 1438 |
| LMArena French | 1522 | 1430 |
| LMArena German | 1522 | 1420 |
| LMArena Japanese | 1510 | 1404 |
| LMArena Korean | 1526 | 1371 |
| LMArena Russian | 1507 | 1407 |
| LMArena Spanish | 1519 | 1395 |

## Instruction Following

- Claude Opus 5: 79.3 (#5)
- o3: 73.0 (#122)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| LMArena Instruction Following | 1516 | 1368 |
| IFEval | — | 86.9% |

## Long Context

- Claude Opus 5: 46.5 (#20)
- o3: 53.2 (#6)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| LMArena Longer Query | 1516 | 1372 |
| Fiction.LiveBench | — | 88.9% |
| CL-bench | — | 17.8% |

## Writing & Preference

- Claude Opus 5: 79.3 (#2)
- o3: 63.7 (#62)

| Benchmark | Claude Opus 5 | o3 |
|---|---|---|
| LMArena Text | 1506 | 1410 |
| LMArena Creative Writing | 1491 | 1359 |
| EQ-Bench Creative Writing | 2133 | 1676 |
| LMArena Multi-Turn | 1496 | 1405 |
| Short-Story Creative Writing | — | 83.9% |
| WildBench | — | 86.1% |
| EQ-Bench 4 | 1385 | — |

## FAQ

### Is Claude Opus 5 better than o3?

Claude Opus 5 is the stronger model overall, scoring 67.7 to 47.4 on the Noometry Index. o3 costs 2.9× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.

### Which is cheaper, Claude Opus 5 or o3?

o3 is cheaper. It lists at $2 per million input tokens and $8 per million output tokens; Claude Opus 5 lists at $5 and $25.

### Is Claude Opus 5 or o3 better for coding?

Claude Opus 5 scores higher on coding benchmarks: 67.0 versus 47.3 in the Noometry coding category.

### Which has the bigger context window?

Claude Opus 5 does, with 1M tokens against 200K.

### How many benchmarks do Claude Opus 5 and o3 share?

34 benchmarks have published results for both models. Claude Opus 5 has 57 scored results on Noometry and o3 has 63.