Source-backed model intelligence
AI Model Index
Compare AI models using source-backed composite indexes, benchmark evidence, coverage, confidence, provenance, and transparent scoring methodology.
Explore model rankings, shared benchmark evidence, source provenance, and the scoring methodology behind each published index.
Frequently asked questions
What makes these AI model rankings different from a typical leaderboard?
Each composite index combines reviewed benchmark rows from multiple evaluator families while preserving coverage, confidence, source provenance, and ranking eligibility. Models with insufficient evidence stay visible but receive no ranked position, and the reasoning behind every published score remains fully inspectable.
Where does the evidence behind the scores come from?
Evidence comes from public benchmark and leaderboard sources such as Artificial Analysis, LLM Stats, and LiveBench-style evaluators. Rows are reviewed, identity-reconciled, and normalized under a fail-closed policy: anything without a declared production adapter or verified source run is excluded from public rankings.
What do coverage and confidence mean here?
Coverage describes how much shared benchmark evidence supports a ranking, and confidence reflects how reliable that evidence is per model. Both are published alongside scores, so you can see when a position rests on thin evidence instead of treating every rank as equally certain.
Is the site free to use, and is the data auditable?
Yes. Browsing, comparisons, methodology pages, and bounded CSV or NDJSON exports are free, with no account required. Every composite score links back to the contributing benchmark rows, source policies, and formula versions, so results can be audited or reproduced independently.