Valen Labs released FrontierBioML, a benchmark that evaluates coding agents on machine learning tasks in computational biology, and published its findings on the company blog. The release includes a leaderboard ranking agents by mean of per-task mean scores with run counts and 95% intervals, alongside average cost per agent, score-over-time charts, and per-task score breakdowns. The benchmark is intended to measure whether coding agents can carry out ML work in computational biology rather than only general software tasks.
Evaluating agents on machine learning tasks in computational biology.
| Mean of per-task mean scores per agent, with run counts and a 95% interval | ||||
|---|---|---|---|---|
| # | Model | n | Mean score | Avg cost |
|---|---|---|---|---|
| Mean score per agent and task, with run scores |
|---|
Please cite this work as:
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.