Model comparison
Claude Opus 4 vs Claude Opus 5
Claude Opus 5 is the stronger model overall, scoring 67.7 to 42.9 on the Noometry Index.
Last verified . 29 shared benchmarks.
Summary
- They share 29 benchmarks with published results for both. Claude Opus 4 scores higher in 0 categories and Claude Opus 5 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Claude Opus 5 leads 77.1 to 27.4.
- The biggest single-benchmark swing is ARC-AGI-2: 8.6% for Claude Opus 4 and 90.4% for Claude Opus 5.
- Claude Opus 5 is cheaper at $5 / $25 per million input/output tokens, against $15 / $75 for Claude Opus 4.
- Claude Opus 5 accepts more context: 1M tokens versus 200K.
Side by side
| Claude Opus 4 | Claude Opus 5 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 42.9 | 67.7 |
| Released | 2025-05-22 | 2026-07-24 |
| Weights | Proprietary | Proprietary |
| Context window | 200K | 1M |
| Max output | 32K | 128K |
| Input $ / M tokens | $15 | $5 |
| Output $ / M tokens | $75 | $25 |
| Results tracked | 56 | 57 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Opus 5 leads
Claude Opus 4: 47.8 (#55), Claude Opus 5: 67.0 (#5)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| WeirdML | 43.7% | 91.8% |
| LMArena Coding | 1442 | 1533 |
| SWE-bench Verified | 70.7% | — |
| DeepSWE | — | 73.6% |
| FrontierCode | — | 53.4% |
| SWE-bench Verified (bash only) | 67.6% | — |
| Aider Polyglot | 72% | — |
| CursorBench | — | 46.6% |
| LMArena WebDev | — | 1694 |
| FrontierSWE | — | 52% |
| SciCode | — | 56.4% |
| GSO | 6.9% | — |
| ALE-Bench | — | 2,165 |
| AlgoTune | 1.33 | — |
Agentic & Tool Use Claude Opus 5 leads
Claude Opus 4: 34.5 (#39), Claude Opus 5: 55.5 (#1)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| APEX-Agents | — | 65.8% |
| OSWorld 2.0 | — | 31.4% |
| τ²-bench Banking | — | 48.7% |
| Cybench | 38% | — |
| DeepResearch Bench | 46.8% | — |
| PostTrainBench | — | 35% |
| BALROG | — | 63.4% |
| GBAEval | — | 79.6% |
| GDP.pdf | — | 24% |
| LMArena Search | 1127 | — |
| METR Time Horizons | 63.9% | — |
| Vending-Bench 2 | — | 11,182 |
Reasoning Claude Opus 5 leads
Claude Opus 4: 27.4 (#104), Claude Opus 5: 77.1 (#3)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| ARC-AGI-2 | 8.6% | 90.4% |
| SimpleBench | 58.8% | 80.6% |
| ARC-AGI-1 | 35.7% | 97.5% |
| CritPt | 0.3% | 29.1% |
| LMArena Hard Prompts | 1399 | 1525 |
| DTBench | 81.6% | 97.9% |
| LMCA | 37.4% | 64.5% |
| Epoch Capabilities Index | 142.67 | 162.67 |
| Kagi LLM Benchmark | 74.3% | — |
| NYT Connections (extended) | — | 94.3% |
| Chess Puzzles | — | 42% |
| EnigmaEval | 5.6% | — |
| EBR-Bench | — | 45.7% |
| Mystery Game Puzzles | — | 59% |
| Bench to the Future 3 | — | 0.12 |
| ForecastBench | 61.1 | — |
Math Claude Opus 5 leads
Claude Opus 4: 40.2 (#80), Claude Opus 5: 86.1 (#6)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 64.4% | 98.9% |
| LMArena Math | 1391 | 1533 |
| FrontierMath (Tiers 1-3) | — | 85.6% |
| FrontierMath Tier 4 | — | 73.2% |
| ProofBench | — | 99% |
| Omni-MATH | 61.6% | — |
| MATH Level 5 | 85% | — |
| FrontierMath (Feb 2025 set) | 4.5% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |
Knowledge Claude Opus 5 leads
Claude Opus 4: 43.9 (#82), Claude Opus 5: 66.8 (#8)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| GPQA Diamond | 76.3% | 93.9% |
| LMArena Expert | 1384 | 1558 |
| Humanity's Last Exam | 10.7% | — |
| SimpleQA Verified | — | 59.9% |
| MMLU-Pro | 87.5% | — |
| Confabulations | 15.9% | — |
| Vectara Hallucination Rate | 12% | — |
| GPQA (HELM) | 70.8% | — |
Multimodal Claude Opus 5 leads
Claude Opus 4: 30.9 (#101), Claude Opus 5: 51.1 (#6)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| LMArena Vision | 1192 | 1321 |
| GeoBench | 49% | — |
| VPCT | 38% | — |
| Blueprint-Bench 2 | — | 30.4% |
| Furniture Assembly | — | 60.8% |
| LMArena Document | — | 1516 |
Multilingual Claude Opus 5 leads
Claude Opus 4: 49.0 (#133), Claude Opus 5: 58.8 (#3)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| LMArena Non-English | 1362 | 1500 |
| LMArena Chinese | 1386 | 1575 |
| LMArena French | 1370 | 1522 |
| LMArena German | 1391 | 1522 |
| LMArena Japanese | 1332 | 1510 |
| LMArena Korean | 1318 | 1526 |
| LMArena Russian | 1392 | 1507 |
| LMArena Spanish | 1388 | 1519 |
Instruction Following Claude Opus 5 leads
Claude Opus 4: 77.3 (#26), Claude Opus 5: 79.3 (#5)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| LMArena Instruction Following | 1406 | 1516 |
| IFEval | 91.8% | — |
Long Context Claude Opus 5 leads
Claude Opus 4: 39.5 (#167), Claude Opus 5: 46.5 (#20)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| LMArena Longer Query | 1422 | 1516 |
| Fiction.LiveBench | 61.1% | — |
Writing & Preference Claude Opus 5 leads
Claude Opus 4: 61.4 (#85), Claude Opus 5: 79.3 (#2)
| Benchmark | Claude Opus 4 | Claude Opus 5 |
|---|---|---|
| LMArena Text | 1377 | 1506 |
| LMArena Creative Writing | 1386 | 1491 |
| EQ-Bench Creative Writing | 1580 | 2133 |
| LMArena Multi-Turn | 1396 | 1496 |
| Short-Story Creative Writing | 83.6% | — |
| WildBench | 85.2% | — |
| EQ-Bench 4 | — | 1385 |
Frequently asked questions
Is Claude Opus 4 better than Claude Opus 5?
Claude Opus 5 is the stronger model overall, scoring 67.7 to 42.9 on the Noometry Index.
Which is cheaper, Claude Opus 4 or Claude Opus 5?
Claude Opus 5 is cheaper. It lists at $5 per million input tokens and $25 per million output tokens; Claude Opus 4 lists at $15 and $75.
Is Claude Opus 4 or Claude Opus 5 better for coding?
Claude Opus 5 scores higher on coding benchmarks: 67.0 versus 47.8 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 4 and Claude Opus 5 share?
29 benchmarks have published results for both models. Claude Opus 4 has 56 scored results on Noometry and Claude Opus 5 has 57.