I built an Ollama alternative with TurboQuant, model groups, and multi-GPU support

· Dev.to

Read full story at source