Wingman
Wingman is a chatbot tool designed to run Large Language Models (LLMs) locally on both PC and Mac (Intel or Apple Silicon). Built with an e...
Last verified:
What is Wingman?
Wingman 2 is a chatbot application that enables users to run large language models (LLMs) locally on their PC and Mac computers without requiring any coding knowledge or terminal usage. The tool provides an intuitive graphical interface that makes running LLMs approachable for anyone, letting users simply point, click, and converse with AI models.
Key features include support for a wide range of cutting-edge language models like Llama 2, Phi, OpenAI GPT4, Mistral, Yi, and Zephyr. Wingman evaluates model compatibility upfront based on your system's specs to avoid crashes or slow performance, shows which LLMs are compatible with your machine at a glance, and allows browsing and accessing the latest LLMs directly from Hugging Face's model hub without leaving the app. Users can customize system prompts and create templates for different use cases, prompting models like characters or with specific viewpoints.
Wingman is designed for anyone who wants to run LLMs locally on Windows PCs with Nvidia GPUs (or CPU-based inference) or MacOS devices (both Intel and Apple Silicon). It is particularly suited for users concerned about privacy who don't want to share secrets with OpenAI, Google, or other companies, as well as open-source enthusiasts who want to contribute to the project.
The application is completely open-source and free to use. It runs entirely on your own machine without phone-home capabilities or network reliance except for initially downloading models. Wingman can be run fully offline once models are downloaded, making it suitable for offline environments.
The first beta release, called Rooster, is now available. The project is actively maintained with frequent updates planned, including bug fixes and a future roadmap. An API is currently in development but not yet ready for public release.
Wingman pricing
Pricing model: Free
Wingman is open-source and completely free. There are no paid plans, subscription fees, or tiered pricing. Users can download the app from the website and give it a try themselves without any cost.
Wingman pros
- Completely free and open-source
- No coding or terminal knowledge required
- Intuitive graphical user interface
- Runs LLMs locally for privacy
- Supports Windows PC with Nvidia GPUs
- Supports CPU-based inference on PC
- Supports MacOS Intel devices
- Supports MacOS Apple Silicon devices
- Evaluates model compatibility upfront
- Prevents crashes and slow performance
- Browses Hugging Face model hub directly
- Sorts models by Emotion IQ
- Customizable system prompts
- Create prompt templates for different use cases
- Works fully offline after model download
- No network requirements for running models
- Doesn't phone home or share data
- Access to Llama 2, Phi, GPT4, Mistral, Yi, Zephyr
- Active development with frequent updates
- Easy 3-step setup process
Wingman cons
- Only beta release available (Rooster)
- API not ready for public release yet
- No multi-modal prompting currently
- Multi-modal input still internally tested
- Must download models before offline use
- Limited to specific model types listed
- GPU compatibility depends on Nvidia on PC
- May have performance limits on older hardware
- Bug fixes still needed out of the gate
Frequently asked questions about Wingman
How much does it cost?
Wingman is open-source, so it's completely free. You can download it from the website and give it a try yourself without any cost.
What are the system requirements?
Wingman works on Windows PCs and MacOS. On PC, it supports Nvidia GPUs or CPU-based inference. On MacOS, it supports both Intel and Apple Silicon devices.
How do I use it?
Using Wingman is easy: 1) Download the app, 2) Run the installer, 3) Use it. The process requires no coding or terminal knowledge.
Is it secure?
Yes, Wingman runs entirely on your own machine, so you aren't sharing your secrets with OpenAI, Google, or anyone else. The app doesn't phone home or rely on the network, except to initially download models.
How do I help out?
The project is open-source, and the developer would love help. You can visit the GitHub repo to report issues, submit pull requests, and contribute to the project.
How often can we expect updates?
You can expect frequent updates as the developer has a lot more planned for Wingman. There will be bug fixes out of the gate, plus a roadmap coming soon.
Does Wingman require the internet, or can it be run fully offline?
Wingman doesn't require the internet to run local models, so you can use it in an offline environment. Just make sure to download the models you want beforehand so you can use them when you don't have network capability.
Does Wingman have an API?
An API is in development but not ready for public release yet. Stay tuned for updates on when it becomes available.
Does Wingman support multi-modal prompting?
Not currently, but multi-modal input is being tested internally. This feature may be added in future updates.
What models does Wingman support?
Wingman supports a wide range of cutting-edge language models including Llama 2, Phi, OpenAI GPT4, Mistral, Yi, and Zephyr. You can browse and access the latest LLMs directly from Hugging Face's model hub without leaving the Wingman app.