benchgap
Other · model

SWE-2 Flash benchmark scores

As of 2026-11-05, SWE-2 Flash (Other) has measured scores on 7 benchmarks and estimated scores on 20 more.

BenchmarkScoreSource
AA-SciCode55.1%estimated ± 3.5 pp, medium confidence
SciCode54.8%estimated ± 7.6 pp, low confidence
Terminal-Bench 2.189.2%measured
AA Coding Index77.1%estimated ± 2.4 pp, low confidence
SWE-bench Verified87.7%estimated ± 0.8 pp, medium confidence
SWE-bench Pro66.0%estimated ± 7.4 pp, low confidence
AA Terminal-Bench 2.185.9%estimated ± 6.8 pp, medium confidence
CursorBench 3.262.5%estimated ± 1.6 pp, medium confidence
Terminal-Bench 3.025.4%estimated ± 9.1 pp, low confidence
FrontierCode 1.1 Main45.5%measured
FrontierSWE v240.0%estimated ± 8.9 pp, medium confidence
Vals LiveCodeBench86.9%estimated ± 1.9 pp, high confidence
Vals SWE-bench85.6%estimated ± 2.8 pp, high confidence
Terminal-Bench 2.1 (Vals AI)75.8%estimated ± 5.4 pp, low confidence
AA Terminal-Bench 4.019.9%estimated ± 5.5 pp, medium confidence
Bug Hunt Bench30.6%estimated ± 8.4 pp, low confidence
CursorBench 4.041.9%estimated ± 2.7 pp, low confidence
DeepSWE60.1%measured
FrontierCode 2.024.0%measured
PostTrainBench v1.128.8%estimated ± 4.5 pp, high confidence
Terminal-Bench 4.020.0%measured
Vibe Code Bench74.4%estimated ± 5.6 pp, low confidence
ExploitBench34.8%measured
NL2Repo56.1%estimated ± 6.2 pp, medium confidence
ScarfBench69.9%estimated ± 3.8 pp, medium confidence
SEC-Bench Pro55.9%estimated ± 3.9 pp, medium confidence
SWE-Atlas Codebase QnA50.0%measured