LocalAI Releases

Patch release to fix incompatibilities with Qwen 3 coder by updating llama.cpp. Includes backend traces feature and drops remaining bark.cpp code.

Feb 21, 2026

Major release featuring multi-modal realtime conversations, new Voxtral TTS backend, multi-GPU support for Diffusers, and UI improvements with dark/light theme.

Feb 20, 2026

Massive update focused on audio and multimodal capabilities with realtime audio conversations, new music generation UI, and expanded ASR/TTS backends.

Feb 7, 2026

This release focuses on significant architectural improvements and new features for the abogen project, including a massive contribution that enabled a Web UI and EPUB 3 pipeline, alongside enhanced TTS capabilities and metadata persistence.

Feb 6, 2026

Patch release with bugfixes for reasoning detection, openresponses support, and API fixes for LTX-2 video generation, plus Qwen-TTS support.

Jan 23, 2026

The release notes for this version are not included in the provided text; only a link to the external changelog is available.

Jan 22, 2026

Major release introducing Anthropic API support, Open Responses API for agents, video generation UI, and unified GPU backends for simplified acceleration.

Jan 18, 2026

This release belongs to the `yazi` project (a terminal file manager), though the detailed changelog content was not provided in the source text.

Jan 4, 2026

This release belongs to the `yazi` project (a terminal file manager), though the detailed changelog content was not provided in the source text.

Dec 29, 2025

A release focused on stability, resource efficiency, and smarter agent workflows with new memory reclamation features.

Dec 24, 2025

This release focuses on improving the Abogen application within LocalAI, introducing the `uv` package manager for faster installation and adding a queue manager option to override item settings globally. It also includes several bug fixes regarding library loading, pip dependencies, and subtitle processing.

Dec 10, 2025

LocalAI enhances Abogen's subtitle capabilities by supporting non-English languages and introducing spaCy for better sentence segmentation. Users can now pre-download models for offline use and utilize more sophisticated timestamp formats.

Nov 28, 2025

A release improving user experience with universal model import, MCP live streaming, and a complete WebUI overhaul.

Nov 26, 2025

This release adds the ability to generate timed audio from subtitle files (.srt, .ass, .vtt) and embed cover images into M4B files. It also addresses Windows compatibility issues regarding DLL initialization and PyTorch CUDA support.

Nov 18, 2025

This release introduces major infrastructure updates including OpenAI-compatible Batch API and Embeddings support, alongside high-performance KVCache offloading frameworks. It also enhances production orchestration with Prefill/Decode (P/D) support and improves runtime stability and autoscaling.

Nov 9, 2025

A release adding Agentic MCP support, the neutts TTS backend, video generation, and Qwen 3 VL support.

Oct 31, 2025

This release belongs to the `RedditVideoMakerBot` project. It includes a fix to stop CI from closing valid issues and general readme updates.

Oct 28, 2025

The interface has been upgraded from PyQt5 to PyQt6 to ensure better long-term compatibility. The queue manager received several usability improvements, including tooltips, separate file open options, and loading animations.

Oct 27, 2025

This release focuses on bug fixes for reranking models, introduces Kokoro support for L4T devices, and adds multilingual capabilities to the Chatterbox backend. It also expands the model gallery with Whisper Turbo, Qwen image edit, and IBM Granite series models.

Oct 3, 2025

A minor release addressing a bug where option checks were not uniform across different backends, alongside standard dependency and documentation updates.

Sep 20, 2025