llama.cpp nightly burst b10668-b10679 includes Vulkan wrong-token fix
llama.cpp
Ten llama.cpp nightlies landed Aug 28 (b10668-b10679). Most consequential is b10677: a Vulkan backend fix for missing view-alias dependencies in ggml_vk_graph_optimize that reportedly produced wrong tokens under greedy decoding on AMD/NVIDIA Vulkan. Others add M3 Max/M4/M5 FlashAttention-Vector Metal tunings, OpenVINO 2026.3.1 with whisper.cpp and NPU Qwen3.5 support, and Qwen4-Exp graph-split reduction.
Why it matters
correctness fix for a silent wrong-output Vulkan bug
Importance: 2/5
nightly patch cycle, but a real wrong-output correctness fix
Sources
official
llama.cpp releases (b10668-b10679)