Priors Lets Developers Tune AI Rankings
Priors lets users compare models head-to-head, then generates an editable benchmark mix that reflects their preferences instead of relying on a single leaderboard. The resulting rankings, cost charts, and benchmark evidence can be inspected and shared.
Priors makes an important point: model rankings are measurement choices, not objective truth, and small weighting changes can reshuffle closely matched systems.
- –Pairwise judgments are converted into a benchmark mix that can rank the broader model pool.
- –Users can edit weights, require stronger measured coverage, and distinguish estimates from actual benchmark results.
- –Its CLI, API, and MCP support make custom rankings reproducible in developer workflows. [CLI and MCP](https://www.mypriors.com/cli)
- –The tool is especially timely as Artificial Analysis changes its index composition and warns that different mixes produce different orderings. [Artificial Analysis indexes](https://artificialanalysis.ai/indexes/) [Index analysis](https://9to5ai.tech/artificial-analysis-index-v5)
DISCOVERED
1h ago
2026-11-03
PUBLISHED
1h ago
2026-11-03
RELEVANCE
AUTHOR
AdamHoltererer