llama.cpp Releases
684 releases of ggml-org/llama.cpp
- b9738
Fixes a security vulnerability in the server where authentication headers were incorrectly forwarded in CORS proxy requests.
Jun 20, 2026
- b9736
Fixes loading of GLM-5.2 GGUF models by marking DSA indexer tensors as optional, allowing models with partial indexer layers to load correctly as full MLA attention.
Jun 20, 2026
- b9735
Optimized the AMX implementation for Intel Xeon processors to improve quantization performance.
Jun 20, 2026
- b9733
Added configuration toggles to enable F16 precision on Vulkan and NVIDIA adapters in the WebGPU backend.
Jun 20, 2026
- b9732
Refactored the internal communication mechanism between child processes and the router in the server architecture.
Jun 19, 2026
- b9731
Optimized the server's get_token_probabilities function using partial sorting to significantly reduce overhead.
Jun 19, 2026
- b9730
Fixed UTF-8 handling issues on Windows affecting the multimodal module and file operations, alongside CLI fixes.
Jun 19, 2026
- b9729
A cleanup release that removes all internal references to 'webui' terminology from the server codebase.
Jun 19, 2026
- b9728
Enhanced the --api-key-file argument to support comment lines, making configuration files more user-friendly.
Jun 19, 2026
- b9727
A maintenance update that upgrades the third-party HTTP library cpp-httplib to version 0.48.0.
Jun 19, 2026
- b9726
The server component introduces a new --agent argument and removes legacy naming compatibility code related to the web UI.
Jun 19, 2026
- b9724
This release focuses on bug fixes within the multimodal system (mtmd), including build stability improvements and overflow corrections.
Jun 19, 2026
- b9723
Adds support for the Eagle3 draft model for Qwen3.5 and Qwen3.6, along with improvements to hybrid model checkpoint handling. The release maintains the standard suite of cross-platform binaries available in the previous update.
Jun 19, 2026
- b9722
This release addresses a bug in the server component where non-bound `n_discard` values caused issues during context shifting. It also provides standard binary distributions for macOS, Linux, Android, Windows, and openEuler across various hardware architectures and backend support.
Jun 19, 2026
- b9721
A synchronization release for the ggml core library, bringing together various updates and fixes.
Jun 19, 2026
- b9718
Refactors the server's slot selection logic to consolidate functions, improving code maintainability while preserving LCP similarity checks.
Jun 19, 2026
- b9717
Optimizes Power10 CPU matmul performance by removing the requirement for the K dimension to be divisible by kc in Q8/Q4 MMA kernels.
Jun 19, 2026
- b9716
Implements batching support for the InternVL multimodal model, improving efficiency when processing multiple inputs.
Jun 19, 2026
- b9715
Implements the CUDA backend for the COL2IM_1D operation, matching the CPU implementation with optimized index decomposition.
Jun 19, 2026
- b9714
Adds a specific HTTP header to streaming endpoints to prevent Nginx reverse proxies from buffering responses, ensuring smooth streaming.
Jun 19, 2026