AI Squawk the real-time terminal for AI news

ggml-backend : start a new split at an incompatible weight in a… — llama.cpp releases

spec + model: refactor DFlash based speculative decoding methods… — llama.cpp releases

hexagon: improve buffer allocations and mappings (mmap stacking) (#32144) … — llama.cpp releases

Blog post: https://huggingface.co/blog/ggml-org/llama-cpp-v1 Overview llama.cpp v1.0.0 removes the deprecated public API and promotes the NextN/speculative… — llama.cpp releases

Model : allow loading external PLE files … — llama.cpp releases

ggml-zdnn: accept non-zdnn tensors, invalidate compute tensors after… — llama.cpp releases

mtmd: refactoring: remove clip.h, consolidate code into mtmd (#31865) — llama.cpp releases

webgpu : add COUNT_EQUAL support … — llama.cpp releases

Assisted-by: dsh:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF… — llama.cpp releases

moe-cache : compute part of the misses on the CPU --moe-cache-cpu-share F computes a share F of the MoE cache misses on the CPU from the host experts instead… — llama.cpp releases

Only run webgpu CI on my fork Add webgpu only workflow Add WebGPU group_norm kernel clean up resolve merge conflicts parameterize data type; clean up cmake… — llama.cpp releases

hexagon: fuse post-norm residual add into RMS_NORM_MUL … — llama.cpp releases

Assisted-by: Claude — llama.cpp releases

CUDA: fix Maxwell build (#31787) — llama.cpp releases