Benchmarks

FrontierCode

Codingunit % · normalized over [0, 70]

FrontierCode v1.1 (Main set) — 150 agentic software-engineering tasks derived from real open-source pull requests, run and graded by Cognition against held-out unit tests plus weighted code-quality rubrics. Composite score, mean@5.

#ModelSourceScoreNormalized

Score distribution

4 tracked results across the normalization window

070

Score vs. parameters

Open-weights models, log-x params

No open-weights models with disclosed parameter counts have a score here yet.