Last released Oct 7, 2026
Hardware-aware auto-tuner that finds the highest-quality llama.cpp GGUF quantization + minimal GPU offload meeting a target throughput and quality-loss budget.