← back to ideas

ClaudeBench

6.8
devtools profitable added: Thursday September 2026 17:35

A web application designed for benchmark testing and comparison of large language models, focusing on scientific reasoning and problem-solving capabilities, mirroring Anthropic's Terminal-Bench-Science approach. It provides a user-friendly interface for defining custom benchmarks and analyzing model performance, allowing users to assess Claude and other models effectively.

150h
mvp estimate
6.8
viability grade
5
views

technology stack

Python NodeJS PostgreSQL Medium

inspired by

Fable 5.1 vs. Fable 5: Results on a real-world budget