Model comparison
Claude Haiku 4.5 vs Claude Opus 4
Claude Opus 4 is the stronger model overall, scoring 43.1 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 15× less per token, which makes it the better buy when Claude Opus 4's lead doesn't matter for your workload.
Last verified . 38 shared benchmarks.
Summary
- They share 38 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 3 categories and Claude Opus 4 in 7 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Claude Opus 4 leads 27.3 to 15.1.
- The biggest single-benchmark swing is ARC-AGI-1: 47.7% for Claude Haiku 4.5 and 35.7% for Claude Opus 4.
- Claude Haiku 4.5 is cheaper at $1 / $5 per million input/output tokens, against $15 / $75 for Claude Opus 4.
Side by side
| Claude Haiku 4.5 | Claude Opus 4 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Noometry Index | 39.5 | 43.1 |
| Released | 2025-10-15 | 2025-05-22 |
| Weights | Proprietary | Proprietary |
| Context window | 200K | 200K |
| Max output | 64K | 32K |
| Input $ / M tokens | $1 | $15 |
| Output $ / M tokens | $5 | $75 |
| Results tracked | 53 | 56 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Opus 4 leads
Claude Haiku 4.5: 44.0 (#78), Claude Opus 4: 47.2 (#62)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| SWE-bench Verified (bash only) | 66.6% | 67.6% |
| WeirdML | 45.4% | 43.7% |
| LMArena Coding | 1453 | 1442 |
| SWE-bench Verified | — | 70.7% |
| Aider Polyglot | — | 72% |
| LMArena WebDev | 1330 | — |
| SWE-bench Multilingual | 64.7% | — |
| SciCode | 43.3% | — |
| GSO | — | 6.9% |
| ALE-Bench | 653.48 | — |
| AlgoTune | — | 1.33 |
Agentic & Tool Use Claude Opus 4 leads
Claude Haiku 4.5: 33.6 (#52), Claude Opus 4: 34.8 (#42)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| DeepResearch Bench | 45.5% | 46.8% |
| Terminal-Bench | 35.5% | — |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| Cybench | — | 38% |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |
| LMArena Search | — | 1127 |
| METR Time Horizons | — | 63.9% |
| Vending-Bench 2 | 458.89 | — |
Reasoning Claude Opus 4 leads
Claude Haiku 4.5: 15.1 (#320), Claude Opus 4: 27.3 (#121)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| ARC-AGI-2 | 4% | 8.6% |
| ARC-AGI-1 | 47.7% | 35.7% |
| CritPt | 0% | 0.3% |
| LMArena Hard Prompts | 1420 | 1399 |
| DTBench | 73.6% | 81.6% |
| LMCA | 30.9% | 37.4% |
| Epoch Capabilities Index | 142.41 | 142.67 |
| ForecastBench | 61.4 | 61.1 |
| SimpleBench | — | 58.8% |
| Kagi LLM Benchmark | — | 74.3% |
| NYT Connections (extended) | 14.3% | — |
| Chess Puzzles | 8% | — |
| EnigmaEval | — | 5.6% |
Math Claude Haiku 4.5 leads
Claude Haiku 4.5: 44.9 (#78), Claude Opus 4: 42.0 (#86)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 64.4% |
| Omni-MATH | 56.1% | 61.6% |
| LMArena Math | 1396 | 1390 |
| MATH Level 5 | 96.4% | 85% |
| FrontierMath (Feb 2025 set) | 5.9% | 4.5% |
| FrontierMath Tier 4 (v1) | 2.1% | 4.2% |
Knowledge Claude Opus 4 leads
Claude Haiku 4.5: 37.7 (#153), Claude Opus 4: 44.0 (#88)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| GPQA Diamond | 71.2% | 76.3% |
| MMLU-Pro | 77.7% | 87.5% |
| Vectara Hallucination Rate | 9.8% | 12% |
| GPQA (HELM) | 60.5% | 70.8% |
| LMArena Expert | 1442 | 1386 |
| Humanity's Last Exam | — | 10.7% |
| SimpleQA Verified | 13.2% | — |
| Confabulations | — | 15.9% |
Multimodal Claude Opus 4 leads
Claude Haiku 4.5: 26.8 (#118), Claude Opus 4: 31.5 (#106)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| LMArena Vision | — | 1192 |
| GeoBench | — | 49% |
| VPCT | — | 38% |
| Blueprint-Bench 2 | 0% | — |
| LMArena Document | 1420 | — |
Multilingual Claude Haiku 4.5 leads
Claude Haiku 4.5: 49.9 (#129), Claude Opus 4: 48.8 (#138)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| LMArena Non-English | 1377 | 1362 |
| LMArena Chinese | 1417 | 1386 |
| LMArena French | 1408 | 1372 |
| LMArena German | 1375 | 1391 |
| LMArena Japanese | 1339 | 1331 |
| LMArena Korean | 1347 | 1321 |
| LMArena Russian | 1381 | 1392 |
| LMArena Spanish | 1420 | 1389 |
Instruction Following Claude Opus 4 leads
Claude Haiku 4.5: 71.4 (#149), Claude Opus 4: 77.1 (#28)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| IFEval | 80.1% | 91.8% |
| LMArena Instruction Following | 1414 | 1406 |
Long Context Claude Haiku 4.5 leads
Claude Haiku 4.5: 43.6 (#92), Claude Opus 4: 39.6 (#172)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| LMArena Longer Query | 1427 | 1422 |
| Fiction.LiveBench | — | 61.1% |
Writing & Preference Claude Opus 4 leads
Claude Haiku 4.5: 57.9 (#123), Claude Opus 4: 61.2 (#89)
| Benchmark | Claude Haiku 4.5 | Claude Opus 4 |
|---|---|---|
| LMArena Text | 1396 | 1377 |
| LMArena Creative Writing | 1372 | 1387 |
| WildBench | 83.9% | 85.2% |
| LMArena Multi-Turn | 1409 | 1396 |
| Short-Story Creative Writing | — | 83.6% |
| EQ-Bench Creative Writing | — | 1580 |
| EQ-Bench 4 | 1064 | — |
Frequently asked questions
Is Claude Haiku 4.5 better than Claude Opus 4?
Claude Opus 4 is the stronger model overall, scoring 43.1 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 15× less per token, which makes it the better buy when Claude Opus 4's lead doesn't matter for your workload.
Which is cheaper, Claude Haiku 4.5 or Claude Opus 4?
Claude Haiku 4.5 is cheaper. It lists at $1 per million input tokens and $5 per million output tokens; Claude Opus 4 lists at $15 and $75.
Is Claude Haiku 4.5 or Claude Opus 4 better for coding?
Claude Opus 4 scores higher on coding benchmarks: 47.2 versus 44.0 in the Noometry coding category.
Which has the bigger context window?
Both accept 200K tokens.
How many benchmarks do Claude Haiku 4.5 and Claude Opus 4 share?
38 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Claude Opus 4 has 56.