Model comparison
Claude 3.5 Haiku vs Claude Sonnet 4
Claude Sonnet 4 is the stronger model overall, scoring 40.6 to 28.9 on the Noometry Index.
Last verified . 37 shared benchmarks.
Summary
- They share 37 benchmarks with published results for both. Claude 3.5 Haiku scores higher in 2 categories and Claude Sonnet 4 in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Claude Sonnet 4 leads 41.5 to 13.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 4.3% for Claude 3.5 Haiku and 71.1% for Claude Sonnet 4.
Side by side
| Claude 3.5 Haiku | Claude Sonnet 4 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 28.9 | 40.6 |
| Released | 2024-10-22 | 2025-05-22 |
| Weights | Proprietary | Proprietary |
| Context window | — | 200K |
| Max output | — | 64K |
| Input $ / M tokens | — | $3 |
| Output $ / M tokens | — | $15 |
| Results tracked | 49 | 58 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Sonnet 4 leads
Claude 3.5 Haiku: 32.6 (#258), Claude Sonnet 4: 44.0 (#70)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| Aider Polyglot | 28% | 61.3% |
| SciCode | 27.4% | 40% |
| WeirdML | 30.7% | 46.1% |
| LMArena Coding | 1287 | 1414 |
| SWE-bench Verified (bash only) | — | 64.9% |
| GSO | — | 4.9% |
| BigCodeBench Instruct | 46.1% | — |
| LiveBench Coding | 51.4% | — |
| BigCodeBench Complete | 59% | — |
| CadEval | 32% | — |
| ALE-Bench | — | 655.35 |
Agentic & Tool Use Claude Sonnet 4 leads
Claude 3.5 Haiku: 27.7 (#90), Claude Sonnet 4: 38.3 (#29)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| TheAgentCompany | — | 33.1% |
| Cybench | — | 35% |
| DeepResearch Bench | — | 46.6% |
| OSWorld | — | 43.9% |
| BALROG | 19.3% | — |
| METR Time Horizons | — | 62% |
Reasoning Claude Sonnet 4 leads
Claude 3.5 Haiku: 17.1 (#286), Claude Sonnet 4: 23.0 (#169)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| CritPt | 0% | 0.3% |
| LMArena Hard Prompts | 1251 | 1372 |
| DTBench | 56.7% | 77.1% |
| Epoch Capabilities Index | 127.15 | 141.69 |
| ARC-AGI-2 | — | 5.9% |
| SimpleBench | — | 45.5% |
| Kagi LLM Benchmark | — | 73% |
| ARC-AGI-1 | — | 40% |
| EnigmaEval | — | 3.1% |
| LiveBench Reasoning | 28.1% | — |
| LiveBench Data Analysis | 48.5% | — |
| LMCA | — | 29% |
| ForecastBench | — | 60.2 |
| LiveBench | 43.5% | — |
Math Claude Sonnet 4 leads
Claude 3.5 Haiku: 13.7 (#294), Claude Sonnet 4: 41.5 (#71)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 4.3% | 71.1% |
| Omni-MATH | 22.4% | 60.2% |
| LMArena Math | 1244 | 1375 |
| MATH Level 5 | 46.4% | 84.4% |
| FrontierMath (Feb 2025 set) | 0.3% | 4.1% |
| LiveBench Math | 35.5% | — |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge Claude Sonnet 4 leads
Claude 3.5 Haiku: 18.7 (#276), Claude Sonnet 4: 41.6 (#99)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| GPQA Diamond | 38.1% | 79.2% |
| MMLU-Pro | 60.5% | 84.3% |
| Confabulations | 36.7% | 13.2% |
| GPQA (HELM) | 36.3% | 70.6% |
| LMArena Expert | 1208 | 1371 |
| Humanity's Last Exam | — | 7.8% |
| Vectara Hallucination Rate | — | 10.3% |
| MMLU | 74.3% | — |
Multimodal Too close to call
Claude 3.5 Haiku: 26.3 (#113), Claude Sonnet 4: 25.6 (#116)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| LMArena Vision | 1092 | 1191 |
| GeoBench | 34% | 37% |
| VPCT | — | 34% |
| MindCube | — | 44.8% |
Multilingual Claude Sonnet 4 leads
Claude 3.5 Haiku: 40.2 (#213), Claude Sonnet 4: 46.9 (#151)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| LMArena Non-English | 1238 | 1333 |
| LMArena Chinese | 1229 | 1351 |
| LMArena French | 1262 | 1364 |
| LMArena German | 1237 | 1331 |
| LMArena Japanese | 1175 | 1301 |
| LMArena Korean | 1172 | 1293 |
| LMArena Russian | 1253 | 1355 |
| LMArena Spanish | 1261 | 1356 |
Instruction Following Claude Sonnet 4 leads
Claude 3.5 Haiku: 63.1 (#229), Claude Sonnet 4: 71.8 (#141)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| IFEval | 79.2% | 84% |
| LMArena Instruction Following | 1241 | 1375 |
| LiveBench Instruction Following | 61.9% | — |
Long Context Claude 3.5 Haiku leads
Claude 3.5 Haiku: 38.3 (#195), Claude Sonnet 4: 33.6 (#254)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| LMArena Longer Query | 1261 | 1398 |
| Fiction.LiveBench | — | 46.9% |
Writing & Preference Claude Sonnet 4 leads
Claude 3.5 Haiku: 42.9 (#228), Claude Sonnet 4: 57.3 (#127)
| Benchmark | Claude 3.5 Haiku | Claude Sonnet 4 |
|---|---|---|
| LMArena Text | 1255 | 1351 |
| LMArena Creative Writing | 1233 | 1345 |
| Short-Story Creative Writing | 73.5% | 81.4% |
| EQ-Bench Creative Writing | 1146 | 1483 |
| WildBench | 76% | 83.8% |
| LMArena Multi-Turn | 1265 | 1376 |
| LiveBench Language | 35.4% | — |
Frequently asked questions
Is Claude 3.5 Haiku better than Claude Sonnet 4?
Claude Sonnet 4 is the stronger model overall, scoring 40.6 to 28.9 on the Noometry Index.
Is Claude 3.5 Haiku or Claude Sonnet 4 better for coding?
Claude Sonnet 4 scores higher on coding benchmarks: 44.0 versus 32.6 in the Noometry coding category.
How many benchmarks do Claude 3.5 Haiku and Claude Sonnet 4 share?
37 benchmarks have published results for both models. Claude 3.5 Haiku has 49 scored results on Noometry and Claude Sonnet 4 has 58.