Today, we’re announcing Arcee AI’s Series B, led by Vista Equity Partners, Cambium Capital, and Emergence Capital, with participation from AI10 Ventures, Hitachi, IAG, M12, P7, and Wipro. The round values Arcee at more than $1 billion.

This financing marks a new chapter for Arcee. We started with a simple conviction: the organizations building with AI should be able to understand, adapt, deploy, and own the models at the center of their work.

That conviction requires more than making existing models smaller and easier to serve: it means building capable open-weight foundation models from the ground up, and building the products and infrastructure that let developers, enterprises, researchers, and public institutions put those models to work on their own terms.

With this funding, we’ll accelerate the next generation of the Trinity model family, expand our work with the U.S. Department of Energy and its national laboratories, and build a new generation of products around open models.

Getting to the frontier

Less than a year ago, we made the decision to build our own family of frontier open-weight models.

That decision became Trinity.

Over the course of six months, we scaled all the way from a 4.5B dense model to Trinity Large, a 400B MoE model. Trinity Large gave us and the broader open-model community a highly capable, permissively licensed model developed end to end in the United States, the first of its kind since Meta stopped releasing the Llama models in early 2025.

Building it also reinforced our core belief that frontier-scale AI does not require frontier-scale waste. Our entire 2025 model lineup, including Trinity Large, was built for approximately $20 million. That includes salaries, compute, data, infrastructure and operations.

Constraints forced us to be precise. We had to build an efficient training pipeline, make extreme sparsity work at scale, focus our post-training on the most impactful developer workloads and carry that same discipline into inference. The result is a model family designed to be highly effective but also dramatically more economical to use.

What comes next

We’ve already shown that we can be competitive with the frontier. Today, this Series B gives us the resources we need to push that frontier forward by strategically investing in the following three areas.

First, we’ll scale our model program and finish building the next generation of Trinity, already in training. We’ll create highly capable models across scales, from efficient models that can run on phones and laptops to frontier systems built for complex scientific and developer workloads.

Second, we’ll expand our work with the Department of Energy and the national laboratories. Genesis-Science-1 is an important starting point for demonstrating how open models can support research institutions working on some of the world’s hardest problems on their own terms.

Third, we’ll grow the product suite around our models. Open weights are the foundation on which we stand, but developers also need the tools to customize, evaluate, deploy, and operate those models reliably. We’ll make the entire AI stack, from pretraining a model to serving in production simpler, faster, and more accessible.

We’re grateful to Vista Equity Partners, Cambium Capital, Emergence Capital, AI10 Ventures, Hitachi, IAG, M12, P7, and Wipro for backing that vision. The combination of financial and strategic partners in this round reflects something we see every day: open models are becoming critical infrastructure for enterprises and institutions around the world.

Most of all, we’re grateful to the team that got us here, and to the developers, researchers, customers, and partners who have tested our models, pushed them, found their limits, and helped us make them better.

The last chapter was about proving that a focused team could build a frontier open-weight model efficiently in the United States.

The next is about building the best open models in the world.