Arcee Series B: $1B+ Valuation After a $20M 400B Model

Arcee AI announced a Series B on September 16, 2026 that values the company at more than $1 billion. Vista Equity Partners, Cambium Capital, and Emergence Capital led. The company's entire 2025 model lineup, including Trinity Large, a 400-billion-parameter mixture-of-experts model, cost about $20 million.

The company blog and the same-day Arcee GlobeNewswire do not disclose the dollar size of the round. They name the valuation, the training bill, and the model.

Participation came from AI10 Ventures, Hitachi, IAG, M12, P7, and Wipro. The byline on the blog is Lucas Atkins, Mark McQuade, and the Arcee AI team. The wire identifies McQuade as co-founder and CEO and Atkins as chief technology officer.

$20 million for a 400B U.S. open-weight model

Over about six months, Arcee says it scaled from a 4.5-billion-parameter dense model to Trinity Large. The blog calls Trinity Large a 400B MoE built end to end in the United States, permissively licensed, and "the first of its kind since Meta stopped releasing the Llama models in early 2025."

The company wire adds the sparsity number the blog leaves out: 400 billion total parameters, 13 billion active per token.

The $20 million figure covers the entire 2025 lineup, including Trinity Large. Arcee itemizes salaries, compute, data, infrastructure, and operations. That is the cost the company wants the Series B priced against. It is not a claim about inference spend, and it is not a disclosed check size.

The blog's own read: constraints forced an efficient training pipeline, extreme sparsity at scale, and post-training aimed at developer workloads. "Frontier-scale AI does not require frontier-scale waste" is the line they put on the page.

"Arcee has demonstrated a rare combination of technical execution and capital efficiency, building frontier open-weight models at a pace and cost profile that stands out in the market."

Monti Saroya, Senior Managing Director, Vista Equity Partners — Arcee GlobeNewswire, September 16, 2026

Next Trinity, DOE labs, and a product suite

Arcee lists three uses of proceeds, and the next Trinity generation is already in training. The target range is efficient models that run on phones and laptops through frontier systems for scientific and developer workloads.

Second, expand work with the U.S. Department of Energy and the national laboratories. The blog names Genesis-Science-1 as the starting point. The company wire puts that work under the Genesis Open Models Initiative and says the DOE system includes 17 national laboratories. The pitch is the one open weights are supposed to win: run the model on your own machines, inspect how it was built, adapt it for a scientific discipline.

Third, grow the product suite around the models. Open weights are the base. Developers still need tools to customize, evaluate, deploy, and operate them. That is the commercial layer sitting on top of Trinity.

"Organizations should not have to choose between the capabilities of a frontier model and the ability to control the technology at the center of their work."

Mark McQuade, co-founder and CEO, Arcee AI — company GlobeNewswire, September 16, 2026

The disclosed numbers are the ones that matter: more than $1 billion on the cap table, about $20 million to train the 2025 lineup, 400 billion parameters with 13 billion active, and six months from 4.5B dense to Trinity Large. The round size is not on the primary paper.