Run GPT models locally
Free boilerplate code to run GPT models on your own device. Keep your chats private & control your GPT usage. Built on top of the best open source LLMs.
Proudly built using:
Powered by leading open-source LLMs
Models are served through the Fireworks.ai API. As new open-source models emerge, yo-GPT’s boilerplate makes integrating them simple.
- Models included out of the box:
- Llama 3.3 70B Instruct$0.90/M Tokens - 128k Context
A 70-billion parameter instruction-tuned model designed for general-purpose tasks with a strong balance of performance and cost.
- Llama 3.1 405B Instruct$3.00/M Tokens - 128k Context
A massive 405-billion parameter model optimized for complex instruction following and high-accuracy tasks, offering extended capabilities for advanced use cases.
- mixtral-8x22b Instruct$1.20/M Tokens - 64k Context
A Mixture of Experts model combining eight 22-billion parameter experts, delivering high-quality outputs with efficient performance, particularly well-suited for creative and coding tasks.
- deepseek-r1$8.00/M Tokens - 160k Context
A high-end research-grade model focused on deep reasoning and long-context understanding, ideal for tasks requiring extensive context and precision.
No subscriptions. Only pay for what you use.
Forget monthly fees. With yo-GPT, you pay based on usage and can easily track your conversation costs and total spend.
Make it yours with custom GPTs and settings.
Tailor yo-GPT to fit your workflow. Create custom GPT profiles, fine-tune settings like token limits and temperature, and decide whether to autosave conversations or keep them ephemeral. Total control, exactly how you want it.
Get all set in 4 steps:
- Register with Fireworks.ai
I am not affiliated. I just love their service.
- Download the boilerplate code
- Go through a couple of instructions
- Chat with your GPTs
👋 Hey, it's Sylvain
I'm passionate about making AI and technology more accessible.
I built yo-GPT for people who want to use LLMs on their own terms — without being locked into expensive monthly subscriptions and without sacrificing privacy.
yo-GPT is a free, open-source boilerplate designed to run LLMs locally using the Fireworks API, giving you full control over your data, your costs, and your AI workflow. If you're privacy-conscious, cost-aware, and want flexibility without the bloat, yo-GPT is made for you.
FAQ
Frequently Asked Questions
yo-GPT is a free, open-source boilerplate designed for people who want to run large language models (LLMs) locally, on their own terms.
It's built to work with the Fireworks.ai API, making it easy to integrate the latest open-source models without worrying about complicated setup, privacy concerns, or recurring costs.
yo-GPT is for developers and privacy-conscious users who want full control over their data, expenses, and AI workflows, without being locked into expensive monthly subscriptions or closed platforms.
Yes! yo-GPT is completely free and open-source. You can use, modify, and extend it however you like. Just keep in mind that while yo-GPT is free, any usage of third-party services like the Fireworks API may incur costs based on your usage.
Yes. yo-GPT is designed with privacy in mind. All chat history is stored locally on your device, so you have full control over your data.
For API calls made to Fireworks.ai, prompt and generation data are not stored or logged for open models. That data only exists temporarily in memory during your request.
However, please note that this is accurate as of March 2, 2025, and Fireworks.ai's terms and data policies may change over time. We recommend reviewing their latest documentation for the most up-to-date information.
Yes. yo-GPT uses the Fireworks.ai API to access and run open-source LLMs, so you'll need an account and API key from Fireworks to connect and use those models.
You only pay for what you use—there are no hidden fees or mandatory subscriptions from yo-GPT itself.
Absolutely! yo-GPT is built to make it easy to integrate new open-source models as they become available through Fireworks.
The boilerplate is designed for flexibility, so extending it to support new models is simple.