San Francisco-based Arcee AI has closed a $150 million Series B at a $1.15 billion post-money valuation, cementing its position as one of the best-capitalized American open-weight foundation-model builders. The round, disclosed on September 16, 2026, was co-led by Vista Equity Partners, Cambium Capital and Emergence Capital, with participation from Microsoft's M12, Hitachi, IAG, P7 and Wipro.
Trinity, plus a science model with the DOE
Arcee positions itself as an open-weight counterweight to closed frontier models. Its Trinity family now runs from Trinity Nano at 6B parameters through Trinity Mini at 26B up to Trinity Large Preview, a 400B-total / 13B-active mixture-of-experts model. Alongside the commercial line, the company is training Genesis-Science-1 with the U.S. Department of Energy, aimed at scientific workloads rather than general chat.
The $20M-a-lineup pitch
Arcee's differentiator, aside from open weights, is cost. The company claims its entire 2025 model lineup — including salaries, compute, data, infrastructure and operations — came in at roughly $20 million end-to-end. That figure is designed to reassure enterprise buyers that they are not dependent on billion-dollar training runs to get frontier-class behavior, and to signal to Vista and its co-leads that this platform can be productized.
Where the money goes
The Series B is earmarked for three things: completing the next generation of Trinity across all scales, expanding Genesis-Science-1 with the DOE, and building out the "boring" production platform (customization, evaluation, deployment and monitoring) that turns open weights into an enterprise product. Arcee's raise lands in the same week as TypeSafe AI's $40M seed and Noetive's $41M seed, extending a run of large checks written for AI-native infrastructure companies.
Why it matters
American open-weight labs have been thin on the ground compared with a growing Chinese cohort. A $1B+ valuation for Arcee gives Vista and Emergence a plausible U.S. answer in that stack — with Microsoft-linked capital and a federal customer already in the mix.
Reporting based on coverage from AlphaSignal, GlobeNewswire, TheNextWeb and Fortune.