Benchmark Almanac

BenchFlow

We track 2 benchmarks and 1 arena that BenchFlow made or helped make.

1 in use 2 with public code 3 with a leaderboard here

Arena leaderboards

Top five in each of its 1 arena with three or more rated models, the most rated first.

Benchmarks

Name Kind Released Best score Status
SkillsBench 86-87 containerized expert tasks in 11 domains, run with and without Agent Skills. Verifier tests give pass rate and skill lift. Terminal tasks Feb 2026 66.5% In use
FrontierPhysics Agents do end-to-end physics research on 56 tasks that 42 physicists from 26 institutions built from their own work, across nine domains of theoretical, computational and experimental physics. Each task asks the agent to review the literature, plan, run the research and write a manuscript. Deterministic verifiers and expert rubrics grade whether the research is correct and publishable, over 3 trials. Research science Aug 2026 27.4% Not in cards