Model comparison
Claude Haiku 4.5 vs Claude Sonnet 4
Claude Sonnet 4 is the stronger model overall, scoring 40.8 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 3.0× less per token, which makes it the better buy when Claude Sonnet 4's lead doesn't matter for your workload.
Last verified . 40 shared benchmarks.
Summary
- They share 40 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 6 categories and Claude Sonnet 4 in 4 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in long context, where Claude Haiku 4.5 leads 43.6 to 33.7.
- The biggest single-benchmark swing is MATH Level 5: 96.4% for Claude Haiku 4.5 and 84.4% for Claude Sonnet 4.
- Claude Haiku 4.5 is cheaper at $1 / $5 per million input/output tokens, against $3 / $15 for Claude Sonnet 4.
Side by side
| Claude Haiku 4.5 | Claude Sonnet 4 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 39.5 | 40.8 |
| Released | 2025-10-15 | 2025-05-22 |
| Weights | Proprietary | Proprietary |
| Context window | 200K | 200K |
| Max output | 64K | 64K |
| Input $ / M tokens | $1 | $3 |
| Output $ / M tokens | $5 | $15 |
| Results tracked | 53 | 58 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Claude Haiku 4.5: 44.0 (#78), Claude Sonnet 4: 43.5 (#88)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| SWE-bench Verified (bash only) | 66.6% | 64.9% |
| SciCode | 43.3% | 40% |
| WeirdML | 45.4% | 46.1% |
| LMArena Coding | 1453 | 1414 |
| ALE-Bench | 653.48 | 655.35 |
| Aider Polyglot | — | 61.3% |
| LMArena WebDev | 1330 | — |
| SWE-bench Multilingual | 64.7% | — |
| GSO | — | 4.9% |
Agentic & Tool Use Claude Sonnet 4 leads
Claude Haiku 4.5: 33.6 (#52), Claude Sonnet 4: 38.5 (#31)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| DeepResearch Bench | 45.5% | 46.6% |
| Terminal-Bench | 35.5% | — |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| TheAgentCompany | — | 33.1% |
| Cybench | — | 35% |
| OSWorld | — | 43.9% |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |
| METR Time Horizons | — | 62% |
| Vending-Bench 2 | 458.89 | — |
Reasoning Claude Sonnet 4 leads
Claude Haiku 4.5: 15.1 (#320), Claude Sonnet 4: 22.9 (#187)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| ARC-AGI-2 | 4% | 5.9% |
| ARC-AGI-1 | 47.7% | 40% |
| CritPt | 0% | 0.3% |
| LMArena Hard Prompts | 1420 | 1372 |
| DTBench | 73.6% | 77.1% |
| LMCA | 30.9% | 29% |
| Epoch Capabilities Index | 142.41 | 141.69 |
| ForecastBench | 61.4 | 60.2 |
| SimpleBench | — | 45.5% |
| Kagi LLM Benchmark | — | 73% |
| NYT Connections (extended) | 14.3% | — |
| Chess Puzzles | 8% | — |
| EnigmaEval | — | 3.1% |
Math Claude Haiku 4.5 leads
Claude Haiku 4.5: 44.9 (#78), Claude Sonnet 4: 43.3 (#80)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 71.1% |
| Omni-MATH | 56.1% | 60.2% |
| LMArena Math | 1396 | 1375 |
| MATH Level 5 | 96.4% | 84.4% |
| FrontierMath (Feb 2025 set) | 5.9% | 4.1% |
| FrontierMath Tier 4 (v1) | 2.1% | 0% |
Knowledge Claude Sonnet 4 leads
Claude Haiku 4.5: 37.7 (#153), Claude Sonnet 4: 41.8 (#108)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| GPQA Diamond | 71.2% | 79.2% |
| MMLU-Pro | 77.7% | 84.3% |
| Vectara Hallucination Rate | 9.8% | 10.3% |
| GPQA (HELM) | 60.5% | 70.6% |
| LMArena Expert | 1442 | 1372 |
| Humanity's Last Exam | — | 7.8% |
| SimpleQA Verified | 13.2% | — |
| Confabulations | — | 13.2% |
Multimodal Too close to call
Claude Haiku 4.5: 26.8 (#118), Claude Sonnet 4: 26.2 (#121)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| LMArena Vision | — | 1191 |
| GeoBench | — | 37% |
| VPCT | — | 34% |
| Blueprint-Bench 2 | 0% | — |
| LMArena Document | 1420 | — |
| MindCube | — | 44.8% |
Multilingual Claude Haiku 4.5 leads
Claude Haiku 4.5: 49.9 (#129), Claude Sonnet 4: 46.7 (#156)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| LMArena Non-English | 1377 | 1333 |
| LMArena Chinese | 1417 | 1350 |
| LMArena French | 1408 | 1363 |
| LMArena German | 1375 | 1331 |
| LMArena Japanese | 1339 | 1302 |
| LMArena Korean | 1347 | 1291 |
| LMArena Russian | 1381 | 1355 |
| LMArena Spanish | 1420 | 1357 |
Instruction Following Too close to call
Claude Haiku 4.5: 71.4 (#149), Claude Sonnet 4: 71.7 (#145)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| IFEval | 80.1% | 84% |
| LMArena Instruction Following | 1414 | 1376 |
Long Context Claude Haiku 4.5 leads
Claude Haiku 4.5: 43.6 (#92), Claude Sonnet 4: 33.7 (#259)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| LMArena Longer Query | 1427 | 1398 |
| Fiction.LiveBench | — | 46.9% |
Writing & Preference Too close to call
Claude Haiku 4.5: 57.9 (#123), Claude Sonnet 4: 57.1 (#132)
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| LMArena Text | 1396 | 1351 |
| LMArena Creative Writing | 1372 | 1345 |
| WildBench | 83.9% | 83.8% |
| LMArena Multi-Turn | 1409 | 1376 |
| Short-Story Creative Writing | — | 81.4% |
| EQ-Bench Creative Writing | — | 1483 |
| EQ-Bench 4 | 1064 | — |
Frequently asked questions
Is Claude Haiku 4.5 better than Claude Sonnet 4?
Claude Sonnet 4 is the stronger model overall, scoring 40.8 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 3.0× less per token, which makes it the better buy when Claude Sonnet 4's lead doesn't matter for your workload.
Which is cheaper, Claude Haiku 4.5 or Claude Sonnet 4?
Claude Haiku 4.5 is cheaper. It lists at $1 per million input tokens and $5 per million output tokens; Claude Sonnet 4 lists at $3 and $15.
Is Claude Haiku 4.5 or Claude Sonnet 4 better for coding?
They score almost the same on coding (44.0 vs 43.5); test both on your own repository before choosing.
Which has the bigger context window?
Both accept 200K tokens.
How many benchmarks do Claude Haiku 4.5 and Claude Sonnet 4 share?
40 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Claude Sonnet 4 has 58.