The startup develops decentralized artificial intelligence software that utilizes cryptography and electronic money to enable individuals and organizations to operate their own model training clusters. This platform allows users to contribute to and benefit from AI model development without dependence on centralized systems, promoting broader access to advanced AI capabilities.
Funding
$250K raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
The increasing computational demands of training and running modern AI models create a barrier to entry, concentrating AI capabilities within a few large organizations with access to massive computing resources. Individuals and smaller entities are often excluded from participating in or benefiting directly from AI development.
Solution
EXO Labs offers a distributed computing framework that enables users to pool the resources of their existing devices—phones, laptops, desktops, and specialized hardware—into a unified AI cluster. This platform automatically discovers and connects devices on a peer-to-peer network, dynamically partitioning AI models across them based on available memory and network topology. By leveraging heterogeneous computing resources, EXO Labs allows users to run large AI models without relying on centralized infrastructure, promoting democratization of AI development and inference. The system provides a ChatGPT-compatible API, allowing easy integration with existing applications.
Target Audience
The primary target audience includes AI researchers, developers, and enthusiasts who want to run and experiment with large AI models on their own hardware, as well as organizations seeking to decentralize their AI infrastructure.
Features
- Automatic device discovery and peer-to-peer networking using UDP, manual configuration, or Tailscale.
- Dynamic model partitioning that optimally splits models across devices based on available resources.
- Support for various inference engines, including MLX and Tinygrad, with interoperability between them.
- Compatibility with a wide range of AI models, including LLaMA, Mistral, LlaVA, Qwen, and Deepseek.
- ChatGPT-compatible API for seamless integration with existing applications.
- Ability to run on heterogeneous devices, including CPUs, integrated GPUs, and dedicated GPUs.
- Model storage in a user-configurable directory with support for downloading models from Hugging Face.