LocalAI Releases

LocalAI 4.9.0 focuses on security hardening, performance optimization, and new multimodal capabilities. The release introduces deny-by-default authentication, chat context compression, and a unified UI, while adding support for MiniMax-H3 video generation and Qwen3-TTS on llama.cpp.

Aug 20, 2026

A stability-focused release addressing quality of life issues, fixing LSP panics, reducing MCP timeouts, and resolving OAuth and TCP typing lag issues in server-client mode.

Aug 7, 2026

Maintenance release focusing on dependency updates and minor bug fixes. Includes the addition of the NVIDIA NeMo-Speech.cpp backend and various improvements to the model gallery and documentation.

Aug 7, 2026

Maintenance release addressing multiple bug fixes including GGUF metadata handling, vllm-cpp engine compatibility, and UI stability issues. Also adds new gallery models.

Aug 6, 2026

Exposes granular realtime configuration options for the Phonic plugin and updates core dependencies to align with the latest agents package.

Aug 6, 2026

Updates plugin dependencies to align with the core LiveKit agents package.

Aug 6, 2026

Updates plugin dependencies to align with the core LiveKit agents package.

Aug 6, 2026

Updates plugin dependencies to align with the core LiveKit agents package and the OpenAI plugin.

Aug 6, 2026

Updates plugin dependencies to align with the core LiveKit agents package.

Aug 6, 2026

Major release introducing vllm.cpp (alpha) C++20 engine, 3D generation modality, audio.cpp backend, gallery variants for hardware-specific builds, and significant reliability improvements for distributed mode.

Aug 5, 2026

This release introduces opt-in pacing controls and webhook events for WhatsApp sync, fixes handling for session revocation, and improves self-chat storage logic.

Aug 2, 2026

This release introduces a new `.crushrc` configuration file format that supports shell scripting features, alongside improvements to AWS Bedrock SSO, IlluminOS support, and various agent reliability fixes.

Jul 31, 2026

This release enhances the wacli tool by adding granular control over file sending media types and improving webhook payloads to include locally resolved chat names.

Jul 27, 2026

Scrapling v0.4.12 focuses on smarter crawling with AutoThrottle, introduces CSV and XML exporters, and adds authentication support for the MCP server.

Jul 26, 2026

Crush v0.87.0 introduces full MCP OAuth 2.1 support, Claude Channels integration, and significant performance optimizations for LSP and Bang mode.

Jul 24, 2026

Paperclip v2026.722.0 adds native Windows support for local agents, a new API for run-bound secret access, and the foundational schema for Connections v3.

Jul 22, 2026

Introduces LSP tools for code refactoring, adds a scrollable sidebar with keyboard navigation, allows removable attachments, and improves OAuth workflows and agent reasoning support.

Jul 20, 2026

Fixes layout issues by merging contained fast-layout blocks and declaring requests dependencies.

Jul 20, 2026

Introduces Skill Studio with a three-pane IDE for authoring skills, adds an attention queue and decisions surface, enhances search capabilities, and implements tougher self-healing run recovery.

Jul 20, 2026

Adds quiet mode for long-running sync sessions, preserves quoted message context in replies, fixes Windows auth path issues, and enhances build security with signed binaries and provenance.

Jul 17, 2026