Arcee AI tops $1B valuation with Series B for open-weight models

Arcee AI entered this round with an unusual profile for a U.S. model developer. Founded in 2023 after CEO Mark McQuade left Hugging Face, it first worked on post-training and fine-tuning — helping companies adapt existing models — before betting the company on building its own. McQuade has said Arcee had only about $30 million on hand when it decided to build Trinity, and almost 70% of total capital went into those models, which cost roughly $20 million to develop last year.

The bet was on open weights, where models' core numerical parameters are released publicly so anyone can download, modify and run them. That approach puts Arcee against the U.S. grain: most American developers favor proprietary models, while most Chinese developers have embraced open weights, producing powerful, accessible models that match the most advanced U.S. systems. McQuade's stated concern is control — that organizations should not have to trade frontier capability for the ability to run and govern the technology themselves.

The Series B turns that bet into a funded plan: next-generation Trinity models, continued work with the Department of Energy and 17 national laboratories on the Genesis-Science-1 open-weight model for scientific computing, and further investment in customization and deployment tools. Vista's Monti Saroya framed it as critical infrastructure for enterprises that want AI on their own terms. What that valuation ultimately rests on is whether Arcee can keep building competitive open-weight models at the low cost it has managed so far, and whether enterprises truly prefer controlling their own models over buying from larger, closed-source vendors.