github llama.cpp · 1h ago · inference

llama.cpp b10416 fixes stale UI cache on llama.app

from English

The llama.cpp project released build b10416, addressing a caching issue on the llama.app website where the index.html file was pinned to old builds due to immutable headers.

The server now serves index.html with no-cache, allowing it to revalidate via ETag and preventing stale UI versions.
macOS Apple Silicon (arm64) builds are available for download.
macOS Intel (x64) binaries are included in the release.
iOS XCFramework is provided for mobile integration.
Ubuntu x64, arm64, and s390x CPU builds are released.
Ubuntu Vulkan and OpenVINO variants are available for Linux.
Windows x64 and arm64 builds support CPU, CUDA 12/13, OpenCL, Vulkan, ROCm 7.14, and SYCL.
Android arm64 (CPU) binaries are included.
openEuler x86 and aarch64 builds with ACL Graph support are listed.

Importance 1/3 Trust 1/3 llama.cpp Inference efficiency

Read original

github llama.cpp · just now Live

llama.cpp b10418 adds SYCL host pinned memory support

The llama.cpp project released version b10418, introducing support for host pinned memory to improve SYCL Host-to-Device memory access. This update addresses thread-safety issues within the backend implementation.

github llama.cpp · just now Live

llama.cpp OpenVINO backend adds Qwen3.5 support and memory optimizations

The llama.cpp project has merged significant updates to its OpenVINO backend, primarily enabling support for the Qwen3.5 model family and introducing several memory optimization techniques.

github llama.cpp · 2h ago · 1 view

llama.cpp b10413 adds auto-detection of draft spec type from GGUF metadata

The llama.cpp project released version b10413, which introduces automatic detection of the speculative decoding specification type directly from draft GGUF model metadata.

github llama.cpp · 3h ago · 1 view

llama.cpp b10414 adds TQ2_0 support and optimizes mul_mv kernel

The llama.cpp project released build b10414, introducing support for the GGML_TYPE_TQ2_0 ternary quantization type in the Metal backend. This update also includes optimizations to the contiguous tensor (cont) mul_mv kernel.

github llama.cpp · 4h ago · 1 view

llama.cpp b10412 enables backend sampling for dflash and dspark

The llama.cpp project has released version b10412, which introduces backend sampling support for both the dflash and dspark backends. This update also enables p_min values greater than 0 within the backend sampling logic and adds corresponding safeguards.