Public benchmarks

Benchwright Registry

Open benchmark registry. Browse, run, and compare models — every public run lands here for everyone.

2

benchmarks

2

subsets

2

tasks

2

total runs

2

models

⌘K

All Agentic Math

2 benchmarks · 1 filter applied

Starter & Viridian (agent, colour, 0.5s)

Long-form Pallet Town on NousResearch/pokemon-agent in Game Boy COLOR mode. Frames sampled at 0.5s and encoded to a film. Leave the house, get a starter, reach Viridian City; staged partial credit.

agenticrewardcustom

1 model · 1 impl

An agentic terminal/CLI benchmark where each task is a containerized environment testing an agent's ability to operate a shell to solve real-world problems.

1 model · 1 impl

back to registry