Arcee
Category: Foundation Models / LLMs
U.S. open-intelligence lab building frontier open-weight foundation models (Trinity family) that developers, enterprises, and public institutions can inspect, adapt, deploy, and own. Arcee was founded in 2023. The company is led by Mark McQuade. Based in San Francisco, USA. Team size: 11-50. Total funding raised: $174M+. Latest round: Series B. Key investors include Vista Equity Partners, Cambium Capital, Emergence Capital, Flybridge, M12 (Microsoft's Venture Fund), Samsung Next, Hitachi Ventures, Prosperity7 Ventures, Long Journey Ventures, Wipro, JC2 Ventures, Guidepoint.
AMW Analysis
Arcee operates in the foundation model space, providing an end-to-end platform for training, merging, and deploying domain-specific Small Language Models within secure, in-VPC enterprise environments. Founded in 2023 and based in Miami, the company has raised $50M through a Series A round backed by investors including Emergence Capital, Flybridge, M12, Samsung Next, and Wipro, among others. Its stated value proposition centers on lowering computational costs by up to 90% while maintaining data privacy for organizations building specialized models on their own data.
Recent news flow shows Arcee expanding beyond its SLM platform into large-scale foundation model development. In January 2026, the company released Trinity, a 400B-parameter sparse Mixture-of-Experts model trained for $20M using 2,048 Blackwell B300 GPUs and released under a permanent Apache 2.0 license. In April 2026, Arcee launched a follow-on version, Trinity Large Thinking, built by a 26-person team on the same $20M budget, activating only 13B parameters per token, running 2-3x faster than dense rivals, and priced at $0.90 per million output tokens. Both releases emphasize on-premise availability and open licensing as differentiators.
AMW analysis, generated from 2 tracked news signals.
- Founded
- 2023
- Headquarters
- San Francisco, USA
- Team size
- 11-50
- Total funding
- $174M+
Value proposition
Delivers U.S.-built frontier open-weight models (Trinity) at capital-efficient cost (~$20M for full 2025 lineup including 400B Trinity Large), so organizations get frontier capability with full control to download, adapt, and run on their own infrastructure.
Products and solutions
Trinity model family (Nano 6B, Mini 26B, Large 400B MoE, Large Thinking), Arcee hosted API (OpenAI-compatible) & enterprise deployment, Mergekit (open-source model merging toolkit), NAC (open-source harness for long-running agent work), Genesis-Science-1 (DOE open-weight science model, in development), DistillKit & prior AFM/Spectrum tooling
Unique value
American open-weight frontier lab: capital-efficient end-to-end training of permissively licensed models (Trinity Nano/Mini/Large) plus tools to customize, evaluate, deploy, and operate them—competing with closed labs and Chinese open-weight leaders.
Target customer
Developers, enterprises, and public institutions needing controllable open-weight AI—including regulated industries, research labs (DOE/national labs), and teams deploying agents on owned infrastructure.
Industries served
Financial Services, Legal & Compliance, Healthcare & Life Sciences, Insurance, Telecommunications, Industrial & Energy
Technology advantage
End-to-end U.S. open-weight training stack achieving frontier-scale MoE (Trinity Large 400B/13B active) at ~$20M lineup cost; owns Mergekit; expanding product layer for customize/evaluate/deploy/operate open models.
How they differentiate
Builds frontier open-weight models end-to-end in the U.S. at unusually low cost (Trinity Large 400B MoE for ~$20M total lineup), with open weights and deployment tooling so enterprises and labs can own and run models privately—vs closed frontier APIs and foreign open-weight labs.
Main competitors
Mistral, Cohere, DeepSeek / leading Chinese open-weight labs, Together AI
Key partnerships
U.S. Department of Energy / national labs (Genesis-Science-1 open science model), Bolt.new (Forge partnership for trillion-param-class training data), Hugging Face (multi-million-dollar Hub partnership; exclusive home for models/datasets), AWS (Strategic Collaboration & Marketplace), Microsoft (M12 investment and Azure), Arm, MongoDB, DatologyAI, NVIDIA (Blackwell training case study)
Notable customers
PIMCO, AngelList, Workhuman, Guild, Activeloop, DatologyAI
Major milestones
Merger with Mergekit (Jan 2024), Raised $24M Series A led by Emergence Capital (July 2024), Strategic investment from M12, Samsung Next, Hitachi Ventures (July 2025), Released Trinity Large 400B MoE (~$20M full 2025 lineup cost) (Jan 2026), Released Trinity-Large-Thinking open reasoning model (Apr 2026), Multi-million-dollar Hugging Face Hub partnership (Jun 2026), Announced Genesis-Science-1 with U.S. DOE / national labs (Jul 2026), Bolt.new Forge partnership for open-model training data (Sep 14, 2026), Raised Series B at >$1B valuation led by Vista, Cambium, Emergence (~$150M+) (Sep 16, 2026)
Growth metrics
Raised Series B at >$1B valuation (~$150M+) in Sep 2026 after training Trinity family including 400B MoE for ~$20M; ~30–39 employees; expanding DOE Genesis-Science-1 and next-gen Trinity training.
Market positioning
American open-intelligence lab building frontier open-weight foundation models and the product stack to customize, deploy, and operate them for enterprises, developers, and public institutions (including DOE).
Geographic focus
Primarily North America (HQ San Francisco), competing globally on open-weight frontier models vs Chinese labs and closed U.S. frontier providers; reach via AWS/Azure and DOE/national lab partnerships.
Patents and IP
Proprietary SLM Adaptation algorithms and Spectrum-powered training techniques (specific patent filings not publicly disclosed).
About Mark McQuade
Mark McQuade is the Co-Founder and CEO of Arcee. Prior to founding Arcee, he was an early commercial hire at Hugging Face, where he focused on enterprise growth and the commercialization of open-source AI models. His background includes serving as the Head of Sales at Roboflow, a computer vision startup, and over five years at Amazon Web Services (AWS) as an Account Manager specializing in cloud and AI infrastructure.
Latest news about Arcee
- Arcee AI Raises $150M Series B to Build Open-Weight Enterprise Models
- Arcee has launched Trinity Large Thinking, a 400B-parameter sparse MoE LLM with only 13B active per token, built by a 26‑person team on a $20 M budget. The model is Apache 2.0 licensed, runs 2‑3× fast
- Arcee AI has released Trinity, a 400B sparse Mixture-of-Experts model challenging Meta’s Llama 4. Trained for $20M in 6 months using 2,048 Blackwell B300 GPUs, it delivers frontier reasoning under a p
More Foundation Models / LLMs companies
Official website: https://www.arcee.ai