# AI Model Claim Checker

> Check a claim such as "model X is the best at coding" or "model X is cheaper than model Y" against BenchLM's public evidence. The answer is supported, contradicted, or not established, with the rows it rests on.

## How It Works

1. A decision model reads the sentence into a fixed form: claim type, subject model, comparator, category, benchmark. It never judges the claim.
2. A deterministic check runs the form against the public leaderboards, category lanes, displayable benchmark rows with their verification status, published prices, documented context windows, and measured speed.
3. The verdict states the rows, the basis (like-for-like, directional, or missing), and the caveats. A directional reading never names a winner.

Claim types: leads the overall leaderboard; leads a category leaderboard; scores higher than another model in a category; scores higher than another model on a benchmark; cheaper per token; larger context window; faster streaming; is an open-weight model.

Agents can skip the reader and call `GET /api/tools/claim-check?type=&subject=&comparator=&category=&benchmark=` with model slugs and benchmark keys; `?catalog=1` lists them.

Try the interactive tool: https://benchlm.ai/tools/claim-checker

Canonical page: https://benchlm.ai/tools/claim-checker