Last released May 26, 2026
Ultra-fast local LLM inference with zero-config hardware-optimized speculative decoding.
Supported by