Arcee AI hits $1B+ valuation with Series B for American open-weight models

Arcee AI said it closed a Series B led by Vista Equity Partners, Cambium Capital, and Emergence Capital that values the company at more than $1 billion. The San Francisco lab builds frontier open-weight models (including Trinity Large) and says the capital will accelerate the next Trinity generation, expand work with the U.S. Department of Energy and national labs, and grow products for customizing and running open models.

SOFTWARE desk — same-day company primary that a U.S. open-weight lab crossed a $1B+ valuation while arguing frontier models can be built for far less than hyperscaler burn, with Vista/Cambium/Emergence and strategic corporates backing the next Trinity generation and DOE work.

Participating investors named on the same company pages, still company: AI10 Ventures, Hitachi, IAG, M12, P7, and Wipro. Fortune’s same-day exclusive, by Allie Garfinkle, names M12 as Microsoft’s M12. File the Vista / Cambium / Emergence leads and those named participants as Arcee’s. File the Microsoft label on M12 as Fortune’s. This desk is not inventing an ownership split or a check size. None was printed on the company pages.

Use of proceeds, as the company tells it: accelerate the next generation of the Trinity model family, already in training; expand work with the U.S. Department of Energy and its national laboratories, with Genesis-Science-1 as a starting point and the Genesis Open Models Initiative named on the GlobeNewswire; and grow products that help organizations customize, evaluate, deploy, and operate open models. File that three-part spend plan as Arcee’s. This desk did not sit in a board meeting.

Product context, still company: Arcee says it builds frontier open-weight foundation models from the ground up in the United States. Open-weight means the trained model weights — the numbers that make the model work — are released so others can run and adapt them. The company says the Trinity family scaled in about six months from a 4.5 billion-parameter dense model to Trinity Large, a 400 billion-parameter mixture-of-experts model. Dense here means every part of the model runs on each token. Mixture-of-experts, or MoE, means only some specialist slices of the model fire for each token, which can cut the compute bill. Arcee calls Trinity Large a permissively licensed U.S. model, and says it is the first of its kind since Meta stopped releasing Llama models in early 2025. File that Trinity / 4.5B-to-400B-MoE / U.S. open-weight picture as Arcee’s. This desk did not train or download the model.

Build-cost claim, company — not an independent audit: Arcee says its entire 2025 model lineup, including Trinity Large, was built for approximately $20 million. That figure, the company says, includes salaries, compute, data, infrastructure, and operations. File the approximately-$20 million line as Arcee’s claim. This desk did not see invoices or meter a training run.

Round size: Arcee did not disclose the dollars raised. Fortune reported the Series B at a $1 billion pre-money valuation and said a source familiar with the matter put the raise at at least $150 million. Pre-money means the implied worth before the new cash is counted in. File that dollar amount as Fortune-reported / not company-confirmed. Do not file $150 million as a company PRIMARY figure. Secondary write-ups that flatten Fortune’s “at least $150 million” into a confirmed raise stay off the lead.

Named voices: Mark McQuade, who Fortune identifies as Arcee’s founder (2023) and who already appears on this desk’s Bolt Forge filing as co-founder and CEO; and Monti Saroya, senior managing director at Vista Equity Partners, quoted on the GlobeNewswire. File the names and titles as theirs, via those pages. Their other quotes are color only and stay in Sources. This is not investment advice.

Plain English for the rest of the card: Series B = a growth venture round after Series A. valuation = what investors imply the whole company is worth. open-weight = model weights released so others can run and adapt them. MoE = mixture-of-experts, an architecture that uses only some slices of the model at a time. DOE = U.S. Department of Energy. national lab = a government research lab in that DOE system. pre-money = implied worth before the new cash is counted in. parameter = one of the trained numbers inside a model.

PRIMARY here: Arcee’s 16 Sep 2026 company blog — Tier A PRIMARY company source, the original record — plus the same-day GlobeNewswire company wire, the company syndication. Dual Tier A PRIMARY. Fortune’s same-day exclusive is attributed reporting on the undisclosed raise size, not a second originating newsroom. The company home is product-home context, not a third originating announce. The Series B, the Vista / Cambium / Emergence leads, the more-than-$1 billion valuation, AI10 Ventures / Hitachi / IAG / M12 / P7 / Wipro, the next-Trinity / DOE-and-national-lab / customize-evaluate-deploy-operate spend plan, Genesis-Science-1 and the Genesis Open Models Initiative, the 4.5B-to-Trinity-Large-400B-MoE picture, the approximately-$20 million 2025 lineup claim, and the San Francisco dateline are company-attributed. The at-least-$150 million raise size and the $1 billion pre-money framing stay Fortune-reported. Product and cost claims stay company-attributed — not independently audited here. NOT claimed: a company-confirmed dollar amount raised, ARR, revenue, headcount, independently verified training cost or model quality, that this desk trained Trinity or saw a term sheet, a stock tip, or investment advice. Distinct from the already-filed orbits-2m-pre-seed, noetive-41m-seed, envive-15m-series-a, typesafe-ai-40m-seed, factory-200m-5b-valuation, bolt-forge-open-models, and gensyn-open-1b-auditable.