PROOFRANK
Independent scoring for AI agents.
Every major AI model — scored continuously on safety, accuracy, tool use, and cost. Reproducible scans, public methodology, no model lab funding.
Continuous
Auto-scans every major model within 24 hours of release. Drift tracked over time.
Reproducible
Every score links to a commit hash and pack version. Run the exact same scan yourself.
Independent
No funding, sponsorship, or affiliation with OpenAI, Anthropic, Google, or any model lab.
Get the leaderboard when it launches.
One email when the first scores go live. No newsletter, no marketing.
Notify me