llama.cpp Releases
684 releases of ggml-org/llama.cpp
- v2.0.2
May 31, 2026
- v2.0.1
A minor release focused on fixing CLI release artifact distinction.
May 31, 2026
- v0.5.3
Enhances microsandbox security and functionality with sandbox name limits, clock sync, and new CLI commands.
May 30, 2026
- v3.6.0
Introduces PaddleOCR-VL-1.6, a high-accuracy document parsing model, along with official API SDKs and multi-page TIFF support.
May 28, 2026
- v1.14.1
Patches security vulnerabilities and adds persistent token caching, expanded Dart parsing, and Nix support.
May 27, 2026
- v1.20.16s6-ready
Adds s6-ready support, get-only shares, and music playback features to copyparty, alongside various bug fixes.
May 26, 2026
- v2.1.19
A security patch release addressing an unquoted service path vulnerability (CWE-428) on Windows.
May 18, 2026
- v2026.516.143833
A major update featuring new Linux capture methods (XDG, Pipewire, KWin), Vulkan encoding, and security updates, alongside new packaging support for FreeBSD and OpenSUSE.
May 16, 2026
- v1.13.2
This release focuses on fixing build issues for Linux wheels and WASM stability, alongside adding support for new TTS models like KittenTTS and Supertonic3.
May 13, 2026
- v1.13.1
A maintenance release addressing bugs in NeMo transducer beam search, fixing crashes in speaker diarization, and updating Sense Voice streaming options.
May 8, 2026
- v1.3.0
A significant performance and feature update adding support for Qwen3.5 and Gemma 4 models, improving VRAM usage reporting, adding reproducible runs, and integrating a benchmarking system.
May 5, 2026
- v2.1.18
An update focusing on Windows kernel driver stability (fixing BSODs and packet injection) and improving compatibility with the IVPN client.
May 4, 2026
- v1.20.14autolocalization
This release introduces several UI and configuration improvements, including language-based UI autoselection, dynamic view switching, and support for Nintendo 3DS/Wii audio formats. It also adds environment variable support in configuration files and various network/OS improvements.
Apr 24, 2026
- v3.5.0
This release deeply integrates with the Hugging Face ecosystem, supporting 20 major models via the Transformers backend. It introduces document processing capabilities, including converting office formats to Markdown and exporting OCR results to DOCX, alongside a new browser-based inference SDK.
Apr 21, 2026
- v3.4.1
This patch release focuses on expanding hardware support and fixing dependencies for PaddleOCR-VL. It adds `llama-cpp-server` backend compatibility, AMD and Intel Arc GPU support, resolves dependency issues for Huawei NPU and KunlunXin XPU, and removes page limits in the Docker Compose service.
Apr 14, 2026
- v1.20.13dothidden
This release introduces cosmetic filtering via `.hidden` files, adds thumbnail pregeneration, and improves JXL thumbnail handling on non-musl systems. It also fixes several bugs related to version alerts, permissions, and file system interactions.
Mar 23, 2026
- v1.20.12fix shares in ftp/sftp
This release prioritizes security by fixing a vulnerability in FTP/SFTP shares that allowed unauthorized access to sibling files within a shared folder. It also addresses another minor vulnerability and updates the DOMPurify library.
Mar 11, 2026
- v1.20.11what? nohtml is evolving!
This release addresses a critical security vulnerability where JavaScript could execute in SVG files even with the `nohtml` flag. It introduces a built-in version-checker for security advisories and includes various fixes for upload rules and configuration.
Mar 8, 2026
- v1.2.0
This release introduces a new LoRA-based abliteration engine, broad support for VL (Vision-Language) models, save/resume optimization progress, and new GPU enumeration features.
Feb 14, 2026
- v2.1.7
Critical security release addressing multiple SQL injection vulnerabilities across the backend by introducing parameterized queries and input validation.
Feb 9, 2026