Up to four models
Claude Haiku 5.5 vs Claude Haiku 4.5
On the index (GPT-4 = 100), Claude Haiku 5.5 scores 174 and Claude Haiku 4.5 138. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Claude Haiku 5.5
174158–191
WeightsClosed
ReleasedOct 2026
Price in / out$0.10 / $0.50
Context1M
Core tests taken7 of 9
Claude Haiku 4.5
138122–155
WeightsClosed
ReleasedOct 2025
Price in / out$1.0 / $5.0
Context200K
Core tests taken3 of 9
Claude Haiku 5.5 leads Claude Haiku 4.5 by 36 points, ahead on 3 of 3 shared core tests, and costs 10× less.
Test by testthird-party results
Claude Haiku 5.5Claude Haiku 4.5
Every result as a table
| Test | Claude Haiku 5.5 | Claude Haiku 4.5 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 89.6% | 71.2% |
| SimpleQA Verified | 23.8% | 13.2% |
| Maths | ||
| AIME-style maths (OTIS mock) | 98.9% | 66.7% |
| Chess puzzles | 55.0% | 8.0% |
| Coding | ||
| Code Migration | 44.6% | 7.5% |
| ProgramBench | 2.5% | 0.0% |
| Vibe Code Bench v1.1 | 90.4% | 11.4% |
| Agents | ||
| DTBench | 95.0% | 73.6% |
| LMCA | 51.1% | 30.9% |
| Professional work | ||
| Finance Agent (v2) | 54.1% | 31.0% |
| EMB | 59.0% | 22.7% |
| Legal Research Bench | 43.8% | 10.6% |
| MedCode | 51.5% | 32.7% |
| MedScribe | 89.5% | 85.2% |
| Public Benefits Bench v1.1 | 63.2% | 54.3% |
| SAGE | 45.8% | 31.8% |
| Tax Agent Bench | 31.9% | 5.9% |
| Novel problems | ||
| ARC-AGI-2 | 61.3% | 4.0% |
| ARC-AGI-1 | 89.0% | 47.7% |
Tests at least two of these models have taken. Bold rows are the core tests.
