# Claude Opus 4.5 vs Claude Opus 4.6

> Claude Opus 4.6 is the stronger model overall, scoring 58.2 to 50.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-opus-4-5-vs-claude-opus-4-6
- Last updated: 2026-10-10
- Shared benchmarks: 60

## Summary

- They share 60 benchmarks with published results for both. Claude Opus 4.5 scores higher in 0 categories and Claude Opus 4.6 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where Claude Opus 4.6 leads 63.0 to 38.6.
- The biggest single-benchmark swing is NYT Connections (extended): 52.5% for Claude Opus 4.5 and 92.1% for Claude Opus 4.6.
- Both cost about the same: $5 input and $25 output per million tokens.
- Claude Opus 4.6 accepts more context: 1M tokens versus 200K.

## Snapshot

| | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 50.5 | 58.2 |
| Rank | 47 | 20 |
| Context | 200K | 1M |
| Input $/M | $5 | $5 |
| Output $/M | $25 | $25 |
| Weights | Proprietary | Proprietary |

## Coding

- Claude Opus 4.5: 54.8 (#27)
- Claude Opus 4.6: 57.2 (#20)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| SWE-bench Verified | 76.7% | 78.7% |
| SWE-bench Verified (bash only) | 76.8% | 75.6% |
| LMArena WebDev | 1494 | 1547 |
| SWE-bench Multilingual | 70.7% | 72% |
| GSO | 26.5% | 41.2% |
| WeirdML | 63.7% | 78% |
| LMArena Coding | 1504 | 1536 |
| ALE-Bench | 1,025 | 996.5 |
| AlgoTune | 1.77 | 1.47 |
| FrontierCode | — | 26.6% |

## Agentic & Tool Use

- Claude Opus 4.5: 47.3 (#12)
- Claude Opus 4.6: 51.1 (#4)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| Terminal-Bench | 63.1% | 79.8% |
| Remote Labor Index | 3.8% | 4.2% |
| τ²-bench Banking | 24.7% | 27.3% |
| Cybench | 82% | 93% |
| DeepResearch Bench | 54.8% | 55.3% |
| LMArena Search | 1180 | 1253 |
| METR Time Horizons | 75% | 78.9% |
| Vending-Bench 2 | 4,967 | 8,018 |
| APEX-Agents | — | 46.3% |
| Berkeley Function Calling Leaderboard | 77.5% | — |
| GDPval | 45.5% | — |
| τ²-bench Airline | 84% | — |
| τ²-bench Retail | 79.6% | — |
| τ²-bench Telecom | 92.3% | — |
| OSWorld | 66.3% | — |
| BALROG | 43.5% | — |
| GBAEval | — | 44.1% |

## Reasoning

- Claude Opus 4.5: 42.6 (#51)
- Claude Opus 4.6: 57.8 (#23)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| ARC-AGI-2 | 37.6% | 69.2% |
| SimpleBench | 62% | 67.6% |
| Kagi LLM Benchmark | 80.2% | 83.6% |
| NYT Connections (extended) | 52.5% | 92.1% |
| ARC-AGI-1 | 80% | 94% |
| Chess Puzzles | 12% | 17% |
| EnigmaEval | 11.9% | 7.6% |
| EBR-Bench | 14.3% | 12.7% |
| LMArena Hard Prompts | 1476 | 1527 |
| Mystery Game Puzzles | 22% | 25% |
| DTBench | 89.9% | 91.2% |
| LMCA | 44.5% | 55.8% |
| Epoch Capabilities Index | 150.09 | 155.24 |
| ForecastBench | 60.7 | 60 |
| Thematic Generalization | — | 80.6% |

## Math

- Claude Opus 4.5: 38.6 (#132)
- Claude Opus 4.6: 63.0 (#31)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 34.4% | 66% |
| FrontierMath Tier 4 | 4.9% | 26.8% |
| OTIS Mock AIME 2024-2025 | 86.1% | 94.4% |
| ProofBench | 36% | 50% |
| LMArena Math | 1463 | 1519 |
| FrontierMath (Feb 2025 set) | 20.7% | 40.7% |
| FrontierMath Tier 4 (v1) | 4.2% | 22.9% |
| MathArena Final-Answer Competitions | — | 78.5% |

## Knowledge

- Claude Opus 4.5: 56.5 (#44)
- Claude Opus 4.6: 61.9 (#26)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| GPQA Diamond | 86% | 90.5% |
| Humanity's Last Exam | 25.2% | 34.4% |
| SimpleQA Verified | 45.7% | 47% |
| Vectara Hallucination Rate | 10.9% | 12.2% |
| LMArena Expert | 1487 | 1546 |

## Multimodal

- Claude Opus 4.5: 31.4 (#107)
- Claude Opus 4.6: 37.3 (#74)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| Furniture Assembly | 28.3% | 28.3% |
| LMArena Document | 1462 | 1507 |
| LMArena Vision | — | 1316 |
| GeoBench | 75% | — |
| VPCT | 40% | — |

## Multilingual

- Claude Opus 4.5: 54.3 (#47)
- Claude Opus 4.6: 57.9 (#6)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| LMArena Non-English | 1438 | 1489 |
| LMArena Chinese | 1470 | 1551 |
| LMArena French | 1471 | 1513 |
| LMArena German | 1449 | 1502 |
| LMArena Japanese | 1416 | 1484 |
| LMArena Korean | 1424 | 1464 |
| LMArena Russian | 1447 | 1497 |
| LMArena Spanish | 1458 | 1510 |

## Instruction Following

- Claude Opus 4.5: 77.5 (#19)
- Claude Opus 4.6: 79.5 (#4)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| LMArena Instruction Following | 1478 | 1523 |

## Long Context

- Claude Opus 4.5: 46.5 (#22)
- Claude Opus 4.6: 48.1 (#13)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| CL-bench | 21.1% | 20.7% |
| LMArena Longer Query | 1480 | 1520 |
| CL-bench Life | — | 17% |

## Writing & Preference

- Claude Opus 4.5: 68.1 (#28)
- Claude Opus 4.6: 73.5 (#10)

| Benchmark | Claude Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
| LMArena Text | 1451 | 1503 |
| LMArena Creative Writing | 1445 | 1505 |
| EQ-Bench Creative Writing | 1687 | 1809 |
| LMArena Multi-Turn | 1466 | 1513 |
| EQ-Bench 4 | — | 1223 |

## FAQ

### Is Claude Opus 4.5 better than Claude Opus 4.6?

Claude Opus 4.6 is the stronger model overall, scoring 58.2 to 50.5 on the Noometry Index.

### Which is cheaper, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.6 is cheaper. It lists at $5 per million input tokens and $25 per million output tokens; Claude Opus 4.5 lists at $5 and $25.

### Is Claude Opus 4.5 or Claude Opus 4.6 better for coding?

Claude Opus 4.6 scores higher on coding benchmarks: 57.2 versus 54.8 in the Noometry coding category.

### Which has the bigger context window?

Claude Opus 4.6 does, with 1M tokens against 200K.

### How many benchmarks do Claude Opus 4.5 and Claude Opus 4.6 share?

60 benchmarks have published results for both models. Claude Opus 4.5 has 69 scored results on Noometry and Claude Opus 4.6 has 68.