ProofBench

What ProofBench measures

Archived benchmark. Historical results combine benchmark versions and are excluded from IQ. Use the separately versioned ProofBench 1.1 for current comparisons.

Reading the cost view

ProofBench publishes no per-run cost, so this chart falls back to each model's estimated effective cost per 1M I/O tokens (a runtime estimate from measured or imputed token usage, not a published task cost). Explicit reasoning/nonreasoning siblings are connected where both are public; every other model is a standalone point. Up and to the left is better.

Reading the completion-time view

Each point is a public model's ProofBench score against its estimated completion time (time to last token) for the input and output lengths set by the sliders. Timing comes from the site's per-model speed measurements, not from the benchmark run, so every point is an estimate and labels carry a ~ prefix. The newest generation with a published result is shown by default. Up and to the left is better.

Reading the score bars

Bars rank models by the source-backed benchmark value used for this chart. Longer bars indicate higher published scores.

Data sources