LocalAI
Drop-in OpenAI replacement
Ettore Di Giacinto
LocalAI is the free, Open Source OpenAI alternative. LocalAI act as a drop-in replacement REST API that's compatible with OpenAI API specifications for local inferencing.
It allows you to run LLMs, generate images, audio locally with consumer grade hardware, supporting multiple model families and architectures.
ā ļø Note Before running a model, make sure your device has enough free RAM to support it. Attempting to run a model that exceeds your available memory could cause your device to crash or become unresponsive. Always check the model requirements before downloading or starting it.
This release introduces powerful new features and improvements.
Key highlights include:
- š¤ Agentic MCP Support with full WebUI integration - Build AI agents that use real tools like web search and code execution, fully OpenAI-compatible
- šļø New neutts TTS Backend - Generate natural, high-quality speech with low-latency audio powered by Neuphonic
- š¼ļø WebUI enhancements - Faster, cleaner UI with real-time updates and full YAML model control
- š¬ Long-Text TTS Chunking for Chatterbox - Generate natural-sounding long-form audio by intelligently splitting text
- š§© Advanced Agent Controls - Fine-tune agent behavior with new options for retries, reasoning, and re-evaluation
- šø New Video Creation Endpoint - OpenAI-compatible /v1/videos endpoint for text-to-video generation
- š Fuzzy Gallery Search - Find models even with typos
- š§ Qwen 3 VL Support - Support for Qwen 3 VL with llama.cpp/gguf models
- Enhanced Whisper compatibility - Now supported on various CPU variants to prevent crashes
- Multiple bug fixes for improved stability and OpenAI API compliance
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and new features:
- Added support for multilingual capabilities in Chatterbox
- Improved reranking models for the llama.cpp backend
- Added support for L4T devices in Kokoro
- Introduced new models to the gallery, including Qwen image edit and IBM Granite variants
- Enhanced chat completion and function calling
- Improved backend runtime capability detection
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and bug fixes:
- Improved process management and shutdown procedures
- Enhanced chat completion and function calling
- Added support for new models in the gallery
- Improved backend runtime capability detection
- Enhanced support for various GPU architectures
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and bug fixes:
- Improved process management and shutdown procedures
- Enhanced P2P worker functionality
- Refined chat completion and function calling
- Added support for new models in the gallery
- Improved backend runtime capability detection
- Enhanced support for various GPU architectures
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and new features:
- New backend support for MLX, including audio and visual language models
- Added WAN support for video generation
- New CPU and MPS version of the diffusers backend for image generation without a GPU
- WebUI now allows downloading model configurations and manual model refresh
- Added stop button for running backends in WebUI
- Models can now be imported and edited via the WebUI
- Improved Whisper backend with integrated Voice Activity Detection
- New LocalAI Launcher App (Alpha) for easier installation and management
- Fixed various issues related to AMD graphics cards, macOS compatibility, and CUDA detection
- Enhanced support for macOS, including whisper, diffusers, llama.cpp, MLX, and stable-diffusion.cpp
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and new features:
- WebUI improvements: now size can be set during image generation
- New backends: KittenTTS, kokoro, and dia are now available as backends and models can be installed directly from the gallery
- Support for reasoning effort in the OpenAI chat completion
- Diffusers backend now available for l4t images and devices
- Backends can now be sideloaded from the system
ā ļø Note: New backends need to be warmed up during the first call to download the model files.
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and new features:
- Support for installing backends from local paths
- Improved responsiveness of documentation tables
- Updates to various dependencies and components
ā ļø Breaking change: Intel GPU images have been renamed. If you're using Intel GPU images, please update your configurations accordingly.
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and new features:
- Support for Flux Kontext and Flux krea for image editing
- Improved backend download reliability
- Bug fixes for Intel GPU images and container images
ā ļø Breaking change: Intel GPU images have been renamed. Please check the full release notes for details.
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release introduces object detection capabilities and includes several improvements:
- New API for fast object detection (install the 'rfdetr-base' model to use)
- Improved backend download reliability with defined mirrors
- Various bug fixes across container images, backends, and installation scripts
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release introduces a new modular architecture, making LocalAI lighter and more flexible:
- Backends are now separate from the core binary, allowing for easier updates and customization
- Automatic backend installation based on your hardware (CPU, NVIDIA, AMD, Intel)
- New CLI commands for managing backends
- Enhanced realtime audio APIs with speech started/stopped events
- Intel GPU acceleration for Whisper transcriptions
- Over 50 new models added to the gallery
ā ļø Note: After upgrading, you may need to manually install required backends for existing models from the WebUI or CLI.
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release includes several improvements and new features:
- Automatically install missing backends along with models
- Support for Gemma 3n models (text generation only)
- Meta packages in backend galleries for easier GPU-specific setup
- Improved backend gallery with descriptions and re-ordering
- Various bug fixes and dependency updates
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This new release of LocalAI comes with major improvements and new features.
Key highlights include:
- Backend Gallery: Install and remove backends on the fly with API-driven customization
- Audio Support: Upload audio, PDFs, or text files directly in the UI
- Realtime API: WebSocket support compatible with OpenAI clients for chat apps and agents
- Reasoning UI: Visual thinking indicators for smart models during inference
- Dynamic VRAM Handling: Smarter GPU usage with automatic layer offloading
- 50+ New Models: Huge model gallery update including Qwen3, reasoning models, and multimodal LLMs
- Experimental Video Generation: New video generation endpoint
Full release notes can be found at https://github.com/mudler/LocalAI/releases
New features and improvements include:
- Support for Qwen3 model family
- Experimental video generation endpoint
- Default images are now slimmer without extra Python libraries
- FFmpeg is now included in all images
Full release notes can be found at https://github.com/mudler/LocalAI/releases
This release brings exciting updates to LocalAI:
- New LocalAI logo and branding
- SYCL support added for stablediffusion.cpp
- WebUI enhancements for improved user experience
- Support for the Lumina model family for image generation
- Bug fixes, including issues related to LOCALAI_SINGLE_ACTIVE_BACKEND
Full release notes can be found at https://github.com/mudler/LocalAI/releases
