BioBench Atlas
Atlas/Models/claude-opus-4-6

Exact model identity

Claude Opus 4.6

Anthropic · version status: reported

Similar model names are never merged automatically. Evaluation membership and numeric result rows only reference the exact ID claude-opus-4-6.

Evaluation settings

BenchmarkWorkRun / groupScope
BioMysteryBenchEvaluating Claude's bioinformatics research capabilities with BioMysteryBenchbiomysterybench-official-run
biomysterybench-v8-full-five-episodes-protocol		full · n=99
BioMysteryBenchEvaluating Claude's bioinformatics research capabilities with BioMysteryBenchbiomysterybench-v8-human-difficult
biomysterybench-v8-human-difficult-five-episodes		subset · n=23
BioMysteryBenchEvaluating Claude's bioinformatics research capabilities with BioMysteryBenchbiomysterybench-v8-human-solvable
biomysterybench-v8-human-solvable-five-episodes		subset · n=76
CompBioBenchAgentic systems are adept at solving well-scoped, verifiable problems in computational biologycompbiobench-nonagentic-baselines
compbiobench-v1-nonagentic-api-three-calls-no-files		full · n=100
LAB-Bench FigQAClaude Sonnet 4.6 System Cardlab-bench-figqa-crop-tool
lab-bench-figqa-sonnet46-adaptive-max-crop-tool-five-runs		track
LAB-Bench FigQAClaude Sonnet 4.6 System Cardlab-bench-figqa-no-tools
lab-bench-figqa-sonnet46-adaptive-max-no-tools-five-runs		track
SpatialBenchSpatialBench 159-evaluation repository snapshotspatialbench-repo-159-mini-swe-agent
spatialbench-repo-159-mini-swe-agent		full · n=159

Numeric results

BenchmarkWorkRun / groupMetricValue
BioMysteryBenchEvaluating Claude's bioinformatics research capabilities with BioMysteryBenchbiomysterybench-v8-human-difficult
biomysterybench-v8-human-difficult-five-episodes		Accuracy — human-difficult23.5 percent
BioMysteryBenchEvaluating Claude's bioinformatics research capabilities with BioMysteryBenchbiomysterybench-v8-human-solvable
biomysterybench-v8-human-solvable-five-episodes		Accuracy — human-solvable77.4 percent
CompBioBenchAgentic systems are adept at solving well-scoped, verifiable problems in computational biologycompbiobench-nonagentic-baselines
compbiobench-v1-nonagentic-api-three-calls-no-files		Accuracy3.7 percent
LAB-Bench FigQAClaude Sonnet 4.6 System Cardlab-bench-figqa-crop-tool
lab-bench-figqa-sonnet46-adaptive-max-crop-tool-five-runs		FigQA score78.3 percent
LAB-Bench FigQAClaude Sonnet 4.6 System Cardlab-bench-figqa-no-tools
lab-bench-figqa-sonnet46-adaptive-max-no-tools-five-runs		FigQA score58 percent
SpatialBenchSpatialBench 159-evaluation repository snapshotspatialbench-repo-159-mini-swe-agent
spatialbench-repo-159-mini-swe-agent		Accuracy52.83 percent
SpatialBenchSpatialBench 159-evaluation repository snapshotspatialbench-repo-159-mini-swe-agent
spatialbench-repo-159-mini-swe-agent		Cost0.8456 USD per evaluation
SpatialBenchSpatialBench 159-evaluation repository snapshotspatialbench-repo-159-mini-swe-agent
spatialbench-repo-159-mini-swe-agent		Duration609.15 seconds per evaluation