Calibrated on 193 published benchmarks

Reasoning models you can run

Models that think step by step before answering. They write more, so speed matters more.