llama.cpp Releases
684 releases of ggml-org/llama.cpp
- b10085
Fixes Qwen3-VL vision position embedding interpolation by switching from align_corners=False to align_corners=True, aligning with the reference implementation and fixing grounding coordinate scaling issues.
Jul 22, 2026
- b10084
Adds a safety check for tensor types when reusing Hexagon (DSP) descriptors to prevent errors or crashes associated with mismatched tensor configurations.
Jul 22, 2026
- b10083
Adds a sqrt_softplus function to the top-k Mixture of Experts (MoE) implementation for dsv4, enhancing activation function capabilities.
Jul 22, 2026
- b10082
Adds a warning that triggers once when a specific weight type has no available KleidiAI kernel, improving debuggability and user awareness of hardware acceleration limits.
Jul 22, 2026
- b10081
Fixes a bug where draft repositories were not resolving to their requested sidecar models when using the --spec-type draft-dflash flag, ensuring speculative decoding works as expected.
Jul 22, 2026
- b10080
Updates server error handling to return HTTP 400 instead of 500 for validation errors involving the X-Conversation-Id header, and fixes the UI to properly filter the backend_sampling placeholder.
Jul 22, 2026
- b10079
Fixes a bug in the server component where it failed to properly handle a null llama_context pointer, preventing potential crashes during request processing.
Jul 22, 2026
- b10078
Refactors Vulkan queue handling to use per-instance mutexes and unique handles, integrating VK_KHR_internally_synchronized_queues to bypass host-side locking when driver-side synchronization is supported.
Jul 21, 2026
- b10077
Adds the missing GGML_BACKEND_DL_IMPL macro invocation for the OpenVINO backend to fix issue #25586, ensuring consistent backend initialization across the codebase.
Jul 21, 2026
- b10076
CUDA optimization to vectorize get_rows operation for int4 types.
Jul 21, 2026
- v1.17.0
Major release of Repomix (not llama.cpp) with custom file processors, security improvements, and MCP enhancements.
Jul 21, 2026
- b10075
Hexagon backend adds CLAMP operation support.
Jul 20, 2026
- b10069
OpenCL improvements for Adreno GPU support including broadcast and view offsets.
Jul 20, 2026
- b10068
DFlash model update for proper handling of injected K/V cache.
Jul 18, 2026
- b10067
Fix for quantization of DeepSeek-V4's routing table tensor.
Jul 18, 2026
- b10066
OpenCL update to load and use a specific kernel from binary kernel library.
Jul 17, 2026
- b10064
OpenCL optimization for improved memory access patterns in q4_K quantization.
Jul 17, 2026
- b10063
General sync release for ggml library updates across multiple platforms.
Jul 17, 2026
- b10061
Bug fix for test suite to prevent NaN values in sentinel tensors.
Jul 17, 2026
- b10059
Changes the default implementation for Hadamard matrix multiplication to use a CPU routine.
Jul 17, 2026