Model comparison
Claude Opus 5 vs Claude Sonnet 4.6
Claude Opus 5 is the stronger model overall, scoring 67.8 to 50.3 on the Noometry Index. Claude Sonnet 4.6 costs 1.7× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.
Last verified . 46 shared benchmarks.
Summary
- They share 46 benchmarks with published results for both. Claude Opus 5 scores higher in 10 categories and Claude Sonnet 4.6 in 0 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where Claude Opus 5 leads 86.2 to 52.9.
- The biggest single-benchmark swing is ProofBench: 99% for Claude Opus 5 and 45% for Claude Sonnet 4.6.
- Claude Sonnet 4.6 is cheaper at $3 / $15 per million input/output tokens, against $5 / $25 for Claude Opus 5.
Side by side
| Claude Opus 5 | Claude Sonnet 4.6 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 67.8 | 50.3 |
| Released | 2026-07-24 | 2026-02-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 1M |
| Max output | 128K | 128K |
| Input $ / M tokens | $5 | $3 |
| Output $ / M tokens | $25 | $15 |
| Results tracked | 57 | 57 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Opus 5 leads
Claude Opus 5: 67.5 (#5), Claude Sonnet 4.6: 46.3 (#67)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| DeepSWE | 73.6% | 29.9% |
| FrontierCode | 53.4% | 24.3% |
| LMArena WebDev | 1691 | 1522 |
| SciCode | 56.4% | 46.8% |
| WeirdML | 91.8% | 66.1% |
| LMArena Coding | 1534 | 1504 |
| ALE-Bench | 2,165 | 1,327 |
| SWE-bench Verified | — | 75.2% |
| CursorBench | 46.6% | — |
| FrontierSWE | 52% | — |
Agentic & Tool Use Claude Opus 5 leads
Claude Opus 5: 55.6 (#1), Claude Sonnet 4.6: 39.1 (#28)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| APEX-Agents | 65.8% | 43% |
| OSWorld 2.0 | 31.4% | 9.3% |
| GBAEval | 79.6% | 48.8% |
| GDP.pdf | 24% | 18% |
| Vending-Bench 2 | 11,182 | 7,204 |
| Terminal-Bench | — | 53.4% |
| τ²-bench Banking | 48.7% | — |
| DeepResearch Bench | — | 54.9% |
| OSWorld | — | 72.1% |
| PostTrainBench | 35% | — |
| BALROG | 63.4% | — |
| ExploitBench | — | 23.6% |
| LMArena Search | — | 1221 |
Reasoning Claude Opus 5 leads
Claude Opus 5: 77.2 (#4), Claude Sonnet 4.6: 46.1 (#45)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| ARC-AGI-2 | 90.4% | 60.4% |
| NYT Connections (extended) | 94.3% | 80.9% |
| ARC-AGI-1 | 97.5% | 86.5% |
| CritPt | 29.1% | 3.1% |
| Chess Puzzles | 42% | 13% |
| LMArena Hard Prompts | 1526 | 1484 |
| Mystery Game Puzzles | 59% | 16% |
| DTBench | 97.9% | 89.9% |
| LMCA | 64.5% | 46.5% |
| Epoch Capabilities Index | 162.78 | 152.24 |
| SimpleBench | 80.6% | — |
| Thematic Generalization | — | 76.3% |
| EBR-Bench | 45.7% | — |
| Bench to the Future 3 | 0.12 | — |
| ForecastBench | — | 62 |
Math Claude Opus 5 leads
Claude Opus 5: 86.2 (#8), Claude Sonnet 4.6: 52.9 (#49)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.9% | 85.8% |
| ProofBench | 99% | 45% |
| LMArena Math | 1531 | 1462 |
| FrontierMath (Tiers 1-3) | 85.6% | — |
| FrontierMath Tier 4 | 73.2% | — |
| FrontierMath (Feb 2025 set) | — | 32.4% |
| FrontierMath Tier 4 (v1) | — | 8.3% |
Knowledge Claude Opus 5 leads
Claude Opus 5: 66.8 (#9), Claude Sonnet 4.6: 51.7 (#65)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| GPQA Diamond | 93.9% | 87.4% |
| SimpleQA Verified | 59.9% | 35.5% |
| LMArena Expert | 1557 | 1500 |
| Vectara Hallucination Rate | — | 10.6% |
Multimodal Claude Opus 5 leads
Claude Opus 5: 50.8 (#8), Claude Sonnet 4.6: 38.0 (#68)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| LMArena Vision | 1319 | 1283 |
| Blueprint-Bench 2 | 30.4% | 6.7% |
| LMArena Document | 1516 | 1482 |
| Furniture Assembly | 60.8% | — |
Multilingual Claude Opus 5 leads
Claude Opus 5: 58.8 (#4), Claude Sonnet 4.6: 54.4 (#41)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| LMArena Non-English | 1501 | 1440 |
| LMArena Chinese | 1574 | 1491 |
| LMArena French | 1519 | 1465 |
| LMArena German | 1524 | 1428 |
| LMArena Japanese | 1516 | 1420 |
| LMArena Korean | 1521 | 1411 |
| LMArena Russian | 1507 | 1440 |
| LMArena Spanish | 1519 | 1464 |
Instruction Following Claude Opus 5 leads
Claude Opus 5: 79.2 (#7), Claude Sonnet 4.6: 77.4 (#25)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| LMArena Instruction Following | 1517 | 1475 |
Long Context Claude Opus 5 leads
Claude Opus 5: 46.5 (#21), Claude Sonnet 4.6: 45.3 (#44)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| LMArena Longer Query | 1515 | 1479 |
Writing & Preference Claude Opus 5 leads
Claude Opus 5: 79.2 (#1), Claude Sonnet 4.6: 70.2 (#22)
| Benchmark | Claude Opus 5 | Claude Sonnet 4.6 |
|---|---|---|
| LMArena Text | 1507 | 1458 |
| LMArena Creative Writing | 1491 | 1435 |
| EQ-Bench Creative Writing | 2133 | 1810 |
| EQ-Bench 4 | 1385 | 1207 |
| LMArena Multi-Turn | 1499 | 1464 |
Frequently asked questions
Is Claude Opus 5 better than Claude Sonnet 4.6?
Claude Opus 5 is the stronger model overall, scoring 67.8 to 50.3 on the Noometry Index. Claude Sonnet 4.6 costs 1.7× 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 Claude Sonnet 4.6?
Claude Sonnet 4.6 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; Claude Opus 5 lists at $5 and $25.
Is Claude Opus 5 or Claude Sonnet 4.6 better for coding?
Claude Opus 5 scores higher on coding benchmarks: 67.5 versus 46.3 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do Claude Opus 5 and Claude Sonnet 4.6 share?
46 benchmarks have published results for both models. Claude Opus 5 has 57 scored results on Noometry and Claude Sonnet 4.6 has 57.