Model comparison
Claude Sonnet 4 vs Claude Sonnet 5
Claude Sonnet 5 is the stronger model overall, scoring 54.6 to 40.8 on the Noometry Index.
Last verified . 31 shared benchmarks.
Summary
- They share 31 benchmarks with published results for both. Claude Sonnet 4 scores higher in 0 categories and Claude Sonnet 5 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Claude Sonnet 5 leads 49.1 to 22.9.
- The biggest single-benchmark swing is GSO: 4.9% for Claude Sonnet 4 and 37.3% for Claude Sonnet 5.
- Claude Sonnet 5 is cheaper at $2 / $10 per million input/output tokens, against $3 / $15 for Claude Sonnet 4.
- Claude Sonnet 5 accepts more context: 1M tokens versus 200K.
Side by side
| Claude Sonnet 4 | Claude Sonnet 5 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 40.8 | 54.6 |
| Released | 2025-05-22 | 2026-06-29 |
| Weights | Proprietary | Proprietary |
| Context window | 200K | 1M |
| Max output | 64K | 128K |
| Input $ / M tokens | $3 | $2 |
| Output $ / M tokens | $15 | $10 |
| Results tracked | 58 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Sonnet 5 leads
Claude Sonnet 4: 43.5 (#88), Claude Sonnet 5: 55.5 (#26)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| SciCode | 40% | 54.3% |
| GSO | 4.9% | 37.3% |
| WeirdML | 46.1% | 68.8% |
| LMArena Coding | 1414 | 1483 |
| ALE-Bench | 655.35 | 1,463 |
| DeepSWE | — | 53.8% |
| FrontierCode | — | 42.7% |
| SWE-bench Verified (bash only) | 64.9% | — |
| Aider Polyglot | 61.3% | — |
| CursorBench | — | 34.1% |
| LMArena WebDev | — | 1541 |
Agentic & Tool Use Claude Sonnet 5 leads
Claude Sonnet 4: 38.5 (#31), Claude Sonnet 5: 42.8 (#18)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| APEX-Agents | — | 54.5% |
| TheAgentCompany | 33.1% | — |
| Cybench | 35% | — |
| DeepResearch Bench | 46.6% | — |
| OSWorld | 43.9% | — |
| GBAEval | — | 65.3% |
| LMArena Search | — | 1194 |
| METR Time Horizons | 62% | — |
| Vending-Bench 2 | — | 6,378 |
Reasoning Claude Sonnet 5 leads
Claude Sonnet 4: 22.9 (#187), Claude Sonnet 5: 49.1 (#39)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| SimpleBench | 45.5% | 60.6% |
| CritPt | 0.3% | 16.9% |
| LMArena Hard Prompts | 1372 | 1461 |
| DTBench | 77.1% | 92.5% |
| LMCA | 29% | 50% |
| Epoch Capabilities Index | 141.69 | 156.21 |
| ForecastBench | 60.2 | 61.1 |
| ARC-AGI-2 | 5.9% | — |
| Kagi LLM Benchmark | 73% | — |
| NYT Connections (extended) | — | 75.1% |
| ARC-AGI-1 | 40% | — |
| Chess Puzzles | — | 35% |
| EnigmaEval | 3.1% | — |
| Mystery Game Puzzles | — | 35% |
| Surface Evolver Bench | — | 60% |
| Bench to the Future 3 | — | 0.14 |
Math Claude Sonnet 5 leads
Claude Sonnet 4: 43.3 (#80), Claude Sonnet 5: 66.2 (#27)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 71.1% | 94.7% |
| LMArena Math | 1375 | 1467 |
| FrontierMath (Tiers 1-3) | — | 65.6% |
| FrontierMath Tier 4 | — | 29.3% |
| ProofBench | — | 77% |
| Omni-MATH | 60.2% | — |
| MATH Level 5 | 84.4% | — |
| FrontierMath (Feb 2025 set) | 4.1% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge Claude Sonnet 5 leads
Claude Sonnet 4: 41.8 (#108), Claude Sonnet 5: 55.6 (#47)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| GPQA Diamond | 79.2% | 90.5% |
| LMArena Expert | 1372 | 1490 |
| Humanity's Last Exam | 7.8% | — |
| SimpleQA Verified | — | 33.7% |
| MMLU-Pro | 84.3% | — |
| Confabulations | 13.2% | — |
| Vectara Hallucination Rate | 10.3% | — |
| GPQA (HELM) | 70.6% | — |
Multimodal Claude Sonnet 5 leads
Claude Sonnet 4: 26.2 (#121), Claude Sonnet 5: 42.4 (#31)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| LMArena Vision | 1191 | 1274 |
| GeoBench | 37% | — |
| VPCT | 34% | — |
| Blueprint-Bench 2 | — | 24.9% |
| LMArena Document | — | 1466 |
| MindCube | 44.8% | — |
Multilingual Claude Sonnet 5 leads
Claude Sonnet 4: 46.7 (#156), Claude Sonnet 5: 53.8 (#55)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| LMArena Non-English | 1333 | 1431 |
| LMArena Chinese | 1350 | 1477 |
| LMArena French | 1363 | 1460 |
| LMArena German | 1331 | 1440 |
| LMArena Japanese | 1302 | 1422 |
| LMArena Korean | 1291 | 1411 |
| LMArena Russian | 1355 | 1451 |
| LMArena Spanish | 1357 | 1437 |
Instruction Following Claude Sonnet 5 leads
Claude Sonnet 4: 71.7 (#145), Claude Sonnet 5: 76.3 (#41)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| LMArena Instruction Following | 1376 | 1452 |
| IFEval | 84% | — |
Long Context Claude Sonnet 5 leads
Claude Sonnet 4: 33.7 (#259), Claude Sonnet 5: 44.8 (#55)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| LMArena Longer Query | 1398 | 1463 |
| Fiction.LiveBench | 46.9% | — |
Writing & Preference Claude Sonnet 5 leads
Claude Sonnet 4: 57.1 (#132), Claude Sonnet 5: 69.2 (#25)
| Benchmark | Claude Sonnet 4 | Claude Sonnet 5 |
|---|---|---|
| LMArena Text | 1351 | 1442 |
| LMArena Creative Writing | 1345 | 1416 |
| EQ-Bench Creative Writing | 1483 | 1794 |
| LMArena Multi-Turn | 1376 | 1454 |
| Short-Story Creative Writing | 81.4% | — |
| WildBench | 83.8% | — |
| EQ-Bench 4 | — | 1236 |
Frequently asked questions
Is Claude Sonnet 4 better than Claude Sonnet 5?
Claude Sonnet 5 is the stronger model overall, scoring 54.6 to 40.8 on the Noometry Index.
Which is cheaper, Claude Sonnet 4 or Claude Sonnet 5?
Claude Sonnet 5 is cheaper. It lists at $2 per million input tokens and $10 per million output tokens; Claude Sonnet 4 lists at $3 and $15.
Is Claude Sonnet 4 or Claude Sonnet 5 better for coding?
Claude Sonnet 5 scores higher on coding benchmarks: 55.5 versus 43.5 in the Noometry coding category.
Which has the bigger context window?
Claude Sonnet 5 does, with 1M tokens against 200K.
How many benchmarks do Claude Sonnet 4 and Claude Sonnet 5 share?
31 benchmarks have published results for both models. Claude Sonnet 4 has 58 scored results on Noometry and Claude Sonnet 5 has 51.