benchgap
Other · model

BTL-4 benchmark scores

As of 2026-11-21, BTL-4 (Other) has measured scores on 3 benchmarks and estimated scores on 18 more.

BenchmarkScoreSource
AA-SciCode47.7%estimated ± 3.8 pp, high confidence
GDPval-AA21.4%estimated ± 5.5 pp, medium confidence
LiveCodeBench v666.1%measured
SciCode46.9%estimated ± 3.6 pp, high confidence
τ²-bench results92.9%estimated ± 12.4 pp, low confidence
AA Coding Index57.6%estimated ± 5.3 pp, medium confidence
SWE-bench Verified78.4%measured
BFCL v473.5%measured
SWE-bench Pro56.7%estimated ± 4.5 pp, high confidence
Vals LiveCodeBench79.9%estimated ± 10.0 pp, low confidence
FrontierCode 1.1 Main24.7%estimated ± 4.2 pp, low confidence
Vals SWE-bench74.0%estimated ± 6.6 pp, medium confidence
DeepSWE45.0%estimated ± 12.7 pp, low confidence
Vibe Code Bench26.0%estimated ± 8.4 pp, medium confidence
MCP Atlas73.7%estimated ± 14.2 pp, low confidence
NL2Repo42.4%estimated ± 6.2 pp, medium confidence
PostTrainBench v117.2%estimated ± 5.5 pp, low confidence
React Native Evals78.9%estimated ± 2.7 pp, medium confidence
SWE-Rebench61.6%estimated ± 3.6 pp, medium confidence
LiveCodeBench78.3%estimated ± 14.6 pp, low confidence
PinchBench100.0%estimated ± 5.0 pp, low confidence