Claude Haiku 5.5
Anthropic·Released Oct 7, 2026
Updated Oct 8, 11:55 PM ET
The score is on the board's scale (reference average 50, about 15 points per standard deviation), not a percentage.
By domain
Each domain is scored on the same scale as the overall score. Coverage is the share of the domain's kinds of evaluation with a result; it says how much evidence there is, not how good the model is.
Every result behind the score
The result as the evaluator publishes it, and the reading it gives on the board's scale. A hard evaluation can give a high reading for a modest result; results near 95% or 5% only say "at least" or "at most".
Coding
10 results · 65.0- CyberBench PatchVals AI · max effort89.3%Reading 82.9
- Vibe Code BenchVals AI · max effort90.4%Reading 74.6
- APEX-SWEMercor · max effort55.1%Reading 74.1
- Terminal-Bench 4.0Vals AI · max effort35.4%Reading 68.7
- Vibe Code Bench 1–100Vals AI · max effort20.4%Reading 66.1
- ProgramBenchVals AI · max effortNear the floor2.5%Reading at most 65.7
- Code MigrationVals AI · max effort44.6%Reading 64.0
- Terminal-Bench 4.0 (AA run)Artificial Analysis · high effort21.7%Reading 60.8
- LiveBench CodingLiveBench · xhigh effort63.9%Reading 51.1
- SciCodeArtificial Analysis · high effort48.7%Reading 46.1
Research and reasoning
5 results · 62.2- ProofBenchVals AI · max effort86.0%Reading 76.5
- MysteryMechanismVals AI · max effort35.1%Reading 70.7
- Terminal-Bench ScienceVals AI · max effort10.0%Reading 58.7
- LiveBench ReasoningLiveBench · xhigh effort87.0%Reading 53.7
- Humanity's Last Exam (AA run)Artificial Analysis · high effort37.3%Reading 53.0
Professional work
8 results · 57.8- MedCodeVals AI · max effort51.5%Reading 73.0
- Tax Agent BenchVals AI · max effort31.9%Reading 65.8
- Legal Research BenchVals AI · max effort43.3%Reading 65.7
- APEX-AgentsMercor · max effort56.1%Reading 64.3
- Finance AgentVals AI · max effort54.1%Reading 61.8
- EMBVals AI · max effort59.0%Reading 55.8
- LiveBench Data AnalysisLiveBench · xhigh effort72.8%Reading 50.8
- Harvey Legal Agent BenchmarkVals AI · max effortNear the floor1.3%Reading at most 24.8
Knowledge and accuracy
3 results · 41.7- LiveBench Instruction FollowingLiveBench · xhigh effort66.5%Reading 48.3
- AA-LCRArtificial Analysis · high effort77.3%Reading 44.8
- LiveBench LanguageLiveBench · xhigh effortCapped63.5%Reading 20.0
Shown, not scored
Composite indices, saturated or older evaluations, evaluations still being checked, and vision, writing and multilingual results. They do not move the score.
- IOIVals AI · CodingReference only47.3%
- SRE BenchVals AI · CodingWatching5.7%
- BioMysteryBenchVals AI · Research and reasoningWatching70.4%
- MedScribeVals AI · Professional workWatching89.5%
- AA Intelligence IndexArtificial Analysis · Composite indicesReference only37.8
- LiveBench averageLiveBench · Composite indicesReference only72.1%
- Vals IndexVals AI · Composite indicesReference only54.3%
No result yet
Scored evaluations this model has not taken. A missing result neither adds nor subtracts.
- FrontierMath Tiers 1–3 · Epoch AI
- FrontierMath Tier 4 · Epoch AI
- Chess Puzzles · Epoch AI
- Mystery Game Puzzles · Epoch AI
- τ²-Bench Telecom (AA run) · Artificial Analysis
- τ-Bench Banking (AA run) · Artificial Analysis
- SimpleQA Verified · Epoch AI
- BullshitBench · BullshitBench
- IFBench (AA run) · Artificial Analysis
- Arena Text · LMArena