Benchmarks

AI workflow benchmarks

Public comparisons of Claude, Codex and Qwen on development workflows.

Status · no results published yet

The first benchmark is in preparation

Nothing is published here until a run is complete. When it is, this page will list it with a link to the full write-up.

What every published run will include

Systems: Each tool and model compared, with the exact version used.
Date: When the run was carried out.
Tasks: The workflows each system was asked to complete, and how many.
Scoring: How results were judged — the methodology, written out.
Results: Charts with the underlying numbers in an accessible table.
Raw data: A downloadable file with the raw results.