Maratyszcza

gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI

High-efficiency floating-point neural network inference operators for mobile, server, and Web

Acceleration package for neural networks on multi-core CPUs