List your company

Software Engineering on SWE-Bench Verified (Res%, QR, CS, Mis.)

86.2Resolution Rate (%)Oracle Routing
68.72873.26477.882.336May 16, 2026

Evaluation results

MethodLinks
Oracle Routing2026.0586.2100—0
HyDRA (peak)τ=0.012026.0575.487.512.981
Claude Sonnet 4.6 (strong)2026.0574.286.1082.4
HyDRA (cons.)τ=0.242026.057485.854.170.2
Binary router2026.0573.885.69.175
GPT-5.3 Codex2026.0573.485.257.378
GPT-5.42026.0573.485.225.785
HyDRA (agg.)τ=0.642026.057182.472.519.8
GPT-5.4-mini (cheap)2026.0569.480.578.60
Claude Haiku 4.52026.0569.480.511.769.4