Claude Haiku 4.5: benchmarks and ranking
Claude Haiku 4.5 (Anthropic) — overall rank: #115 / 153, overall score: 25.5 (90% interval 19.3–33.0), leaderboards: 4. Released: 2025-10-01 · Price (in/out): $1 / $5
Rank on each leaderboard
| Leaderboard | Rank | Published value |
|---|---|---|
| Artificial Analysis | #88 / 141 | 16.9 |
| LiveBench | Not rated | — |
| Epoch AI | #106 / 230 | 142.74 |
| ARC Prize (ARC-AGI) | #72 / 95 | 4 |
| Scale AI SEAL | Not rated | — |
| Vals AI | Not rated | — |
| Terminal-Bench | Not rated | — |
| EQ-Bench | #30 / 32 | 1053.4 |
| SimpleBench | Not rated | — |
| BenchLM.ai | Not rated | — |
Domain scores
- Coding: 33.5
- Agents: 22.8
- Reasoning: 24.3
- Math: 44.3
- Knowledge: 50.8
- Writing: 46.0
- Instructions: 26.4
- Chat & EQ: 21.5
- Business: 58.3
- Japanese: 59.8
- Vision: 10.7
- Long context: 5.2