Objective checks (regex / exact tokens), not penning quality. A 135M example is allowed to neglect — that is the measurement. Pick models, afterward run. Estimate uses your final tok/s if we have one.
Speed (tokens/s, sustained decode, suite wall) and accuracy (pass charge on goal tests) from runs in this browser. Numbers remain on this machine. Charts use the latest suite per model.
🏆
Verified Benchmark Certificate & Social Share
Generate and download a verifiable achievement certificate alongside your equipment hardware, highest and sustained tokens/second, and portion your score.
Your Name or Handle
Device & Hardware Information
Write a benchmark in JavaScript
The publishing company is eval()’d in this origin, afterward all inspect runs on the model’s decoded text.
