llama.cpp Releases
609 releases of ggml-org/llama.cpp
- b9914
Fixes a potential crash in the OpenCL AOS (Array Of Structures) reconstruction logic.
Jul 8, 2026
- b9913
Introduces support for Q2_0 quantization, a 2-bit quantization scheme, specifically for the CPU backend, allowing for smaller model sizes.
Jul 8, 2026
- b9912
A minor code cleanup release fixing naming conventions and spacing issues in the speculative decoding module.
Jul 8, 2026
- b9911
Implements fused matrix-vector quantization (MMVQ) post-scale operations specifically for NVIDIA FP4 (NVFP4) quantization on CUDA, resulting in performance improvements.
Jul 8, 2026
- b9910
Resolves inconsistencies between how draft models are fitted and loaded in the server, improving the stability of speculative decoding implementations.
Jul 8, 2026
- b9909
Enhances the `/responses` API stream endpoint by providing detailed timing metrics and progress updates to improve observability for users.
Jul 8, 2026
- b9908
Enforces strict RAM limits for prompt caching in the server API to prevent memory overruns and improves eviction logic by allowing the last entry to be removed.
Jul 8, 2026
- b9907
A minor maintenance fix adding a necessary header file (`<fstream>`) to the common library module to prevent potential compilation errors.
Jul 8, 2026
- b9906
Corrects a compilation error where `-ffast-math` was inadvertently disabling Infinity and NaN handling under ROCm 22, which caused build failures in CI.
Jul 8, 2026
- b9905
Addresses a bug related to the quantized key-value cache for dsv4 (DeepSeek V4) to ensure correct state management and consistency.
Jul 8, 2026
- b9904
This release focuses on fixing unit test cases for specific operations (CONT and CPY) within the SYCL backend. It includes build artifacts for all supported platforms and architectures.
Jul 7, 2026
- b9902
Adds support for SYCL operations cross_entropy_loss and cross_entropy_loss_back, expanding SYCL functionality for machine learning tasks.
Jul 7, 2026
- b9901
Optimizes SYCL DMMV (Dequantize-Mul-Mat-Vec) path by setting K_QUANTS_PER_ITERATION to 1, significantly improving GPU utilization and performance on B70 hardware.
Jul 7, 2026
- b9899
Enhances SYCL argsort implementation to support all unit test cases, improving robustness and test coverage of the sorting functionality.
Jul 7, 2026
- b9898
Uses SYCL function calls to fix Abstract-Optimized-Toolchain (AOT) double type issues, improving type handling in SYCL implementations.
Jul 7, 2026
- b9897
Improves SYCL environment variable naming by changing from 'disable' to 'enable' flags for better user experience and more intuitive configuration.
Jul 7, 2026
- b9895
Fixes out-of-bounds read in ngram-map during prompt shrinking operations, along with cleanup cutoff improvements for speculative decoding.
Jul 7, 2026
- b9894
Vulkan fix to check src0 type in GGML_OP_SET_ROWS operation to prevent failures due to unimplemented f16 support.
Jul 7, 2026
- b9893
Focuses on OpenCL flash attention decode performance optimizations with multiple kernel improvements and bug fixes for various quantization formats and model configurations.
Jul 7, 2026
- b9892
Binary release with extensive platform support including macOS, Linux, Android, and Windows. The primary technical change is setting optimal default thread count for PowerPC (PPC) on Linux and AIX systems.
Jul 6, 2026