B.O.B. Leaderboard
Anthropicreleased 7 Oct 2026
Claude Haiku 5.5
#21 of 9524 of 46 benchmarksConsistent
62.8
B.O.B.
Category profile
- Real use (hard to game)2×64
- Reasoning & knowledge49
- Math74
- Coding69
- Agents & real work42
- Independent indexes78
Reality check
Lab tests
58
Real use (5 tests)
64
Lab tests and real-use tests agree. It does 6 points better on tests that are hard to train for than on the usual lab tests.
What counts as real usePrice per 1M tokens (OpenRouter)
$0.10 in · $0.50 out
Cost per task on the AA index
$0.21
Output speed (Artificial Analysis)
237 tok/s
§Receipts
Every benchmark result
“Beats” = the share of tracked models this one beat on that test. The raw score is its best reported setting.
| Benchmark | Raw | Beats | Source |
|---|---|---|---|
| Real use (hard to game) · Real use 64 | |||
| LiveBench 2.0 (real-work suites)as claude-haiku-5.5 | 77.0% | 76%of 39 | LiveBench / Abacus.AI |
| BridgeBench (overall)as Claude Haiku 5.5 | 601 | 68%of 29 | BridgeMind |
| LMArena WebDevas claude-haiku-5.5-high | 1585 | 68%of 82 | LMArena (blind human votes) |
| Vals Indexas anthropic/claude-haiku-5-5 | 54.4% | 61%of 42 | Vals AI (private test sets) |
| LMArena Agent Arenaas Claude Haiku 5.5 (High) | 0.015 | 47%of 50 | LMArena (real agent sessions) |
| Reasoning & knowledge · Reasoning 49 | |||
| Chess Puzzlesas claude-haiku-5-5_max | 55.0% | 83%of 73 | Epoch AI Benchmarking Hub |
| CritPt (physics research)as claude-haiku-5-5_xhigh | 22.6% | 77%of 89 | Epoch AI Benchmarking Hub |
| Blueprint-Bench 2as claude-haiku-5-5_unknown | 33.0% | 60%of 31 | Epoch AI Benchmarking Hub |
| Mystery Game Puzzlesas claude-haiku-5-5_max | 30.0% | 53%of 57 | Epoch AI Benchmarking Hub |
| GPQA Diamondas claude-haiku-5-5_max | 89.6% | 44%of 70 | Epoch AI Benchmarking Hub |
| Humanity's Last Examas claude-haiku-5-5_unknown | 30.1% | 17%of 13 | Epoch AI Benchmarking Hub |
| SimpleQA Verifiedas claude-haiku-5-5_max | 23.8% | 11%of 57 | Epoch AI Benchmarking Hub |
| Math · Math 74 | |||
| ProofBenchas claude-haiku-5-5_max | 89.0% | 88%of 68 | Epoch AI Benchmarking Hub |
| OTIS Mock AIMEas claude-haiku-5-5_xhigh | 98.9% | 79%of 73 | Epoch AI Benchmarking Hub |
| FrontierMath Tiers 1-3as claude-haiku-5-5_max | 75.1% | 70%of 60 | Epoch AI Benchmarking Hub |
| FrontierMath Tier 4as claude-haiku-5-5_max | 46.3% | 59%of 52 | Epoch AI Benchmarking Hub |
| Coding · Coding 69 | |||
| CursorBenchas claude-haiku-5-5_max | 48.4% | 75%of 17 | Epoch AI Benchmarking Hub |
| SciCodeas claude-haiku-5-5_max | 55.0% | 69%of 86 | Epoch AI Benchmarking Hub |
| FrontierCodeas claude-haiku-5-5_max | 46.4% | 68%of 45 | Epoch AI Benchmarking Hub |
| FrontierSWEas claude-haiku-5-5_max | 43.8% | 63%of 20 | Epoch AI Benchmarking Hub |
| Agents & real work · Agents 42 | |||
| APEX-Agentsas claude-haiku-5-5_max | 56.1% | 64%of 48 | Epoch AI Benchmarking Hub |
| Vending-Bench 2as claude-haiku-5-5_unknown | 3602.5% | 20%of 47 | Epoch AI Benchmarking Hub |
| Independent indexes · Indexes 78 | |||
| Epoch Capabilities Indexas Claude Haiku 5.5 | 159.0 | 84%of 81 | Epoch AI |
| Artificial Analysis Intelligence Indexas Claude Haiku 5.5 (max) | 43 | 72%of 54 | Artificial Analysis |