5h ago · llama.cpp releases →
spec + model: refactor DFlash based speculative decoding methods…
Includes: llama.cpp b12375, v1.0.0-b12359-b1b770d: examples : remove explict version from test-cmake (#32191), llama.cpp b12374, llama.cpp b12373, llama.cpp b12370, b12376: Hexagon: add support for IQ2_XXS and NVFP4 (#31954), llama.cpp b12377