Claude Opus 5.5
Anthropic·Released Sep 22, 2026
Updated Oct 8, 21:18 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 · 84.5- ProgramBenchVals AI · max effort18.5%Reading 110.0
- APEX-SWEMercor · max effort67.6%Reading 92.9
- LiveBench CodingLiveBench · xhigh effort77.3%Reading 89.7
- Terminal-Bench 4.0Vals AI · max effort65.2%Reading 87.4
- Terminal-Bench 4.0 (AA run)Artificial Analysis · high effort56.6%Reading 85.3
- Vibe Code Bench 1–100Vals AI · xhigh effort30.4%Reading 83.1
- Code MigrationVals AI · max effort66.6%Reading 81.4
- SciCodeArtificial Analysis · high effort60.4%Reading 78.3
- Vibe Code BenchVals AI · max effort90.3%Reading 74.3
- CyberBench PatchVals AI · max effort85.7%Reading 68.5
Research and reasoning
8 results · 85.6- FrontierMath Tier 4Epoch AI · max effortNear the ceiling95.0%Reading at least 101.0
- Mystery Game PuzzlesEpoch AI · max effort71.0%Reading 94.2
- ProofBenchVals AI · max effortNear the ceiling100.0%Reading at least 89.2
- Terminal-Bench ScienceVals AI · max effort47.1%Reading 86.0
- MysteryMechanismVals AI · max effort49.5%Reading 85.2
- FrontierMath Tiers 1–3Epoch AI · max effort91.2%Reading 83.9
- LiveBench ReasoningLiveBench · xhigh effort93.7%Reading 78.8
- Humanity's Last Exam (AA run)Artificial Analysis · high effort55.6%Reading 77.2
Professional work
8 results · 71.8- APEX-AgentsMercor · max effort73.5%Reading 87.3
- Tax Agent BenchVals AI · max effort45.6%Reading 83.2
- EMBVals AI · max effort75.9%Reading 77.3
- LiveBench Data AnalysisLiveBench · xhigh effort79.8%Reading 73.8
- Legal Research BenchVals AI · max effort50.5%Reading 73.6
- Finance AgentVals AI · max effort58.6%Reading 70.6
- MedCodeVals AI · max effort49.8%Reading 67.1
- Harvey Legal Agent BenchmarkVals AI · max effortCappedNear the floor3.8%Reading at most 24.8
Knowledge and accuracy
5 results · 67.8- SimpleQA VerifiedEpoch AI · max effort72.2%Reading 88.3
- LiveBench LanguageLiveBench · xhigh effort85.5%Reading 75.5
- BullshitBenchBullshitBench · max effort43.6%Reading 65.6
- AA-LCRArtificial Analysis · high effort82.7%Reading 53.9
- LiveBench Instruction FollowingLiveBench · xhigh effort67.0%Reading 50.6
Human preference
1 result · 63.5- Arena TextLMArena · high effort1,503.7Reading 56.4
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 only95.1%
- MirrorCodeEpoch AI · CodingWatching77.4%
- SRE BenchVals AI · CodingWatching33.6%
- BioMysteryBenchVals AI · Research and reasoningWatching79.3%
- OTIS Mock AIMEEpoch AI · Research and reasoningReference only100.0%
- APEX-AccountingMercor · Professional workWatching14.8%
- EBR-benchEpoch AI · Professional workWatching71.4%
- MedScribeVals AI · Professional workWatching91.4%
- GPQA DiamondEpoch AI · Knowledge and accuracyReference only90.6%
- AA Intelligence IndexArtificial Analysis · Composite indicesReference only53.6
- LiveBench averageLiveBench · Composite indicesReference only82.1%
- Vals IndexVals AI · Composite indicesReference only67.0%
- Arena WebDevLMArena · Writing and designReference only1,813.6
No result yet
Scored evaluations this model has not taken. A missing result neither adds nor subtracts.
- Chess Puzzles · Epoch AI
- τ²-Bench Telecom (AA run) · Artificial Analysis
- τ-Bench Banking (AA run) · Artificial Analysis
- IFBench (AA run) · Artificial Analysis