Benchmark

Toolathlon (Tool Decathlon)

Measures: An agent completes 108 long, multi-app tasks (about 20 interactions each) using 32 MCP servers (604 tools) plus local toolkits, starting from realistic app states; a script checks the final state and the score is Pass@1.

Maintainer
HKUST NLP (hkust-nlp)
Introduced
2025
Status
Active
Saturation
Headroom
Typical provenance
Independent
In scoring set
Yes

PaperOfficial leaderboard

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

No results to show.