The inference-performance king, an industry default
PagedAttention doubles throughput — the foundation of most self-hosted inference.
Run LLMs locally with one command
Cloud inference and fine-tuning for open models
One API, hundreds of models
A desktop GUI for local models