News

Startups

Artificial Intelligence

Americas

Oumi Launches World's First Unconditionally Open AI Platform with $10M Seed Funding

Oumi has debuted a collaborative AI laboratory that provides open code, data, and weights to democratize foundation model development.

Oumi has debuted a collaborative AI laboratory that provides open code, data, and weights to democratize foundation model development.

NewDecoded

Published Apr 2, 2026

Apr 2, 2026

4 min read

Image by OUMI

Seattle-based Oumi emerged from stealth on January 29, 2025, unveiling an unconditionally open source AI platform designed for collective development. The Public Benefit Corporation secured $10 million in Seed funding led by Venrock and Obvious Ventures to build what it calls the world's largest open-source laboratory. This initiative brings together experts to ensure the future of artificial intelligence remains transparent and accessible to the global developer community.

Global Collaboration for AI Research

In collaboration with researchers from 13 top-tier universities including MIT and Stanford, the platform offers an end-to-end ecosystem for training and deploying state-of-the-art models. Oumi stands out by providing not just open model weights, but also the underlying code, data, and recipes necessary for true reproducibility. This academic partnership aims to bridge the gap between theoretical research and practical, high-scale application in the real world. CEO and co-founder Manos Koukoumidis, previously of Google Cloud AI, aims to create a Linux moment for the artificial intelligence industry. He believes opening AI to the community is essential for making technology accessible and advancing it responsibly rather than keeping it behind proprietary black boxes. The focus is on fostering an environment where innovation is driven by collective discovery rather than corporate silos.

A Technical Infrastructure Built for Scale

The platform supports foundational model workflows ranging from 10 million to 405 billion parameters. Developers can utilize advanced techniques like SFT and LoRA while integrating with popular inference engines like vLLM. It is designed to run seamlessly across local hardware and major cloud providers such as AWS and GCP, ensuring that hardware limitations do not hinder creativity or progress.

Beyond technical tools, Oumi emphasizes open collaboration to break down talent silos currently found in major tech firms. Co-founder Oussama Elachqar, formerly of Apple, describes the vision as making AI a team sport where global experts can experiment and discover together. This approach is intended to accelerate the speed of discovery in every area of machine learning and safety research.

Since its launch, the project has gained significant momentum, reaching over 9,000 stars on its GitHub page by early 2026. This community growth paved the way for a commercial enterprise platform that automates complex model-building tasks via a natural language interface. More information on their mission and tools can be found at the Oumi website.


Decoded Take

Decoded Take

Decoded Take

Oumi’s entry into the market challenges the growing trend of open-washing where companies release model weights but keep training data and methodologies proprietary. By operating as a Public Benefit Corporation and providing the full stack of code and data, Oumi establishes a new standard for transparency that mirrors the impact of Linux on the software industry. This move signals a significant shift toward a more reproducible AI infrastructure, potentially breaking the monopoly of big tech firms by allowing academic researchers and smaller enterprises to build high-performance models locally without vendor lock-in.

Share this article

Related Articles