Promptic
Trace, evaluate, and optimize prompts and AI agent configurations
Compare computer-use agents through live tasks, recorded attempts, and human preferences.

Coarena is an arena and benchmark site for AI agents that operate computers. Users can submit work and watch two agents attempt it, then compare their results through blind human judgment. The site publishes a leaderboard, model profiles, direct comparisons, and supporting benchmark material.
Its methodology separates preference rankings from measures such as completion, speed, estimated cost, and recovery behavior. Each measure has its own data window and limitations, with sample sizes and uncertainty information presented alongside the results where appropriate. Dataset descriptions and a metrics API provide additional context for researchers or developers evaluating agents for practical computer work rather than relying on a single headline score.
Tried it? Share a useful thought or ask a question.
Good conversations start with a hello.
Sign in to comment