Lemonade
Lemonade helps users discover and run local AI apps by serving optimized LLMs right from their own GPUs and NPUs. Join our discord: https://discord.gg/5xXzkMu8Zk
Last verified:
What is Lemonade?
Lemonade Server is a lightweight, open-source local LLM server that allows users to run and manage multiple AI applications on their local machine. It serves as a server interface using the standard OpenAI API, enabling applications to integrate with local LLMs and easily replace cloud-based LLMs with private and free LLMs running locally on the user's own PC's NPU and GPU. The tool is automatically optimized for AMD hardware including Ryzen AI, Radeon, and Strix Halo PCs.
Key features include omni-modal capabilities with standard endpoints for chat, vision, image generation, image editing, speech generation, and transcription. It offers a one-click Windows GUI installer with Linux and macOS (beta) installers also available. The server implements the standard OpenAI API specification at http://localhost:8000/api/v1, allowing any OpenAI-compatible client library to connect. Lemonade Server supports multiple models simultaneously, auto-detects hardware, and includes a built-in control panel app.
Lemonade is designed for developers, AI enthusiasts, and anyone wanting private local AI without cloud dependencies. It integrates with hundreds of apps out-of-the-box thanks to OpenAI API compatibility, supports Python, C++, Java, C#, Node.js, Go, Ruby, Rust, and PHP clients, and enables context sizes of 64k or more. The <10 MB binary is portable across Windows, Linux, and macOS.
Lemonade pricing
Pricing model: Freemium
Lemonade Server is completely free and open source with no paid plans. There is no telemetry, no strings attached, and users can customize and redistribute to their heart's content. The tool is available as a standalone download with a one-click Windows GUI installer, and installers are also available for Linux and macOS (beta). Users can download models from the built-in model registry or register and install models from Hugging Face at no cost.
Lemonade pros
- Open source with no strings attached and no telemetry
- Fully OpenAI API compatible for easy integration
- Runs completely locally for maximum privacy
- Free to use with no paid tiers or subscriptions
- Automatically optimized for AMD GPU and NPU hardware
- One-click Windows GUI installer for easy setup
- Supports chat, vision, image gen, speech gen, and transcription
- Can run multiple models simultaneously
- Auto-detects available hardware (CPU, GPU, NPU)
- Portable <10 MB binary works on Windows, Linux, macOS
- Supports context sizes of 64k tokens or more
- Works with hundreds of OpenAI-compatible apps out-of-box
- Native C++ backend for high performance
- Built-in control panel app for management
- Community-driven with active Discord support
Lemonade cons
- macOS installer still in beta stage
- Primarily optimized for AMD hardware
- Requires local GPU or NPU for best performance
- api_key still required in client config (though unused)
- Limited to localhost access by default
- NVIDIA GPU support may be less optimized
- Some backends require manual installation
- Lower VRAM on integrated GPUs limits model size
- Learning curve for CLI commands and API endpoints
- Research/reasoning models need explicit labeling
Frequently asked questions about Lemonade
What is Lemonade Server?
Lemonade Server is a server interface that uses the standard OpenAI API, allowing applications to integrate with local LLMs. It helps users discover and run local AI apps by serving optimized LLMs right from their own GPUs and NPUs. This means you can easily replace cloud-based LLMs with private and free LLMs that run locally on your own PC's NPU and GPU.
Is Lemonade Server free to use?
Yes, Lemonade Server is completely open source, free & private with no strings attached and no telemetry. Users can customize and redistribute it to their heart's content. There are no paid plans or subscription tiers.
What operating systems does Lemonade Server support?
Lemonade Server is available as a standalone tool with a one-click Windows GUI installer. Installers are also available for Linux and macOS (beta). The <10 MB binary is portable and can be deployed on any computer running Windows, Linux, or macOS.
What hardware does Lemonade Server work best with?
Lemonade Server is extra optimized for Ryzen AI, Radeon, and Strix Halo PCs. It automatically detects your hardware including CPU, GPU, and NPU. The tool is built by the community and optimized by AMD, working best with AMD GPUs and NPUs.
How do I integrate Lemonade Server with my application?
Since Lemonade Server implements the standard OpenAI API specification, you can use any OpenAI-compatible client library by configuring it to use http://localhost:8000/api/v1 as the base URL. Official client libraries are available for Python, C++, Java, C#, Node.js, Go, Ruby, Rust, and PHP.
What AI capabilities does Lemonade support?
Lemonade is omni-modal with standard endpoints for chat, vision, image generation, image editing, speech generation, and transcription. Point your app at Lemonade and get all these capabilities with standard APIs through a single local service.
How do I install models with Lemonade Server?
You can use the lemonade-server CLI with the pull command to install models ahead of time, or use the run command to pull and run a model. You can also use the POST /v1/pull API endpoint to register and install models from the built-in registry or from Hugging Face.
Can I run multiple models at once?
Yes, Lemonade Server supports running many models at once. The health endpoint shows maximum models per type: 1 LLM/chat model, 1 embedding model, 1 reranking model, 1 transcription model, 1 image model, and 1 TTS model can be loaded simultaneously.
What is the default API endpoint for Lemonade Server?
The default API endpoint is http://localhost:8000/api/v1 for the OpenAI-compatible API. The server listens on port 8000 by default, and the WebSocket port for log streaming and real-time audio transcription is assigned by the OS or can be set via --websocket-port.
How do I check which models are available?
You can use the lemonade-server list command to check all models available. The models endpoint also returns available models, and you can query it programmatically. The server automatically validates available models by querying the /models endpoint.