Find the right starting point.

A C/C++ inference project for exploring how language models run on hardware you control.

An orchestration library for stateful workflows and agents, with documented persistence and human-interruption mechanisms.

A reference research assistant that combines locally hosted language models with configurable web-search steps.

A model-definition library for inference and training across text, vision, audio and multimodal tasks.

A command-line tool and library for configurable evaluations of prompts, models, agents and retrieval applications.

A speech-recognition codebase and model family for transcription, speech translation and language identification.

A compact, historical GPT training codebase for studying the relationship between a model definition and a training loop.

A hosted review workflow for discussing a proposed change, requesting corrections and recording an approval decision.