An Open-Source AI Model Matches the Big Labs on a Shoestring Budget

Developers are racing to test it, and benchmarks are being argued line by line.

A small research group released an open model that rivals much larger systems on popular benchmarks. Developers on X, Bluesky and Reddit are posting tests and running it on home hardware. Critics say benchmarks don't capture real-world use, but download numbers are climbing fast.