github llama.cpp · 2d ago · inference

llama.cpp b10853 adds Kimi-K3 recurrent-state rollback support

from English

The llama.cpp project has released version b10853, which introduces support for the Kimi-K3 model's recurrent-state rollback mechanism via pull request #28466.

This update provides pre-built binaries and frameworks across multiple platforms and hardware accelerators. Available builds include macOS (Apple Silicon and Intel), iOS, Linux (Ubuntu x64, arm64, s390x with CPU, Vulkan, ROCm 10.0, OpenVINO, and SYCL backends), Android (arm64), Windows (CPU, OpenCL Adreno, CUDA 12/13, Vulkan, OpenVINO, SYCL, ROCm 10.0), and openEuler (x86 and aarch64 with ACL Graph).

The release ensures compatibility with the latest Kimi-K3 architecture while maintaining broad hardware support for local inference.

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

Read original

github llama.cpp · 3h ago · 1 view

llama.cpp b10888 adds Vulkan command-buffer debug labels for GPU profilers

llama.cpp b10886 adds Q1_0 vector intrinsic support for s390x

llama.cpp b10883 fixes Vulkan matrix multiplication and restores Ampere optimizations

llama.cpp b10870 adds f16 B-type matmul pipelines and warp tile tuning for Intel coopmat1

llama.cpp b10871 adds dedicated iq4_xs mat-vec shader for Vulkan