Benchmarks

Public AI Agent Benchmarks

Public task sets and environments that test whether agents can complete work through tools, code, browsers, or multi-turn interaction.

AllBenchmarksLeaderboardsPapersToolsArticles
Showing 13-16 of 16 resourcesPage 2 of 2
Benchmark·Multimodal reasoning
MMMU-Benchmark

MMMU-Pro

A multimodal reasoning benchmark used in frontier model evaluation.

Open resource
Benchmark·Math reasoning
OpenAI

MATH-500

A 500-problem math benchmark built from the held-out MATH test split.

Open resource
Benchmark·Scientific coding
SciCode

SciCode

A benchmark for coding solutions to realistic scientific research problems.

Open resource
Benchmark·Terminal agents
Harbor Framework

Terminal-Bench v2.1

An evaluation for coding agents that use terminal tools to complete tasks.

Open resource
Resource does not match your workflow?

Public benchmarks are references. Your delegated workflow may need a private task set.

Submit an evaluation request