Benchmarks
Agentic benchmarks
These benchmarks measure how well models carry out tasks that take several steps.
We track 379 benchmarks in this topic. Main benchmarks are listed by how many models they cover, most first. Where a benchmark has separate results for a subset of its tasks, those results appear right under it. Click one for its leaderboard.
Showing 101–200 of 379 benchmarks.
- Tau2 Airline20 models
- ACEBench18 models
- mcp-universe-function-call18 models
- Rli18 models
- hal-core-bench-hard17 models
- BFCL-v316 models
- Legal Agent Benchmark16 models
- OSWorld 216 models
- Tau3 Banking15 models
- TheAgentCompany14 models
- Tau-bench13 models
- xbench-$OneMillion-Bench13 models
- BrowserGym WorkArena-L112 models
- hal-gaia12 models
- BrowserGym MiniWoB11 models
- BrowserGym WorkArena-L211 models
- GDPVal Win Rate11 models
- GDPVal Win+Tie11 models
- hal-scienceagentbench11 models
- BFCL10 models
- hal-assistantbench10 models
- hal-tau-bench-airline10 models
- hal-online-mind2web9 models
- hal-usaco9 models
- MCP-Mark9 models
- Finance Agent8 models
- GDPval-AA 2.18 models
- OSWorld 2.1 (partial)8 models
- AndroidWorld_SR7 models
- BrowserGym WebArena7 models
- ClawBench7 models
- SWE Atlas - Codebase QnA7 models