Arcee AI
Arcee AI is a U.S. open-weight model lab building the Apache-2.0 Trinity foundation model family.
Founded
2023Arcee AI is a U.S. open-weight model lab building the Apache-2.0 Trinity foundation model family.
Arcee AI is a U.S.-based open-weight lab offering a sovereign alternative to Chinese open models, with Apache-2.0 licensing that carries no lock-in. Models are downloadable and fully customizable by enterprises for on-premises and edge deployment.
Trinity-Large-Thinking ranked second on the PinchBench benchmark behind Opus-4.6 while costing roughly 96 percent less per million output tokens. The model is positioned as the strongest open-weight reasoning model released by a Western company.
Arcee is betting on American open weights as a sovereign alternative as international labs pivot back toward proprietary models, and plans a trillion-parameter model for frontier reasoning with enterprise transparency. Hugging Face was named the exclusive home for all Arcee models, datasets, and agent traces in June 2026.
The company partner ecosystem spans Nvidia, Intel, ARM, Cisco, AWS, Microsoft, OpenRouter, and Digital Ocean. Arcee positions its open-weight strategy as a way to reclaim the open-source model lead for the United States.
Arcee AI builds the Trinity family of open-weight foundation models under the Apache-2.0 license, designed to run on edge, on-premises, and cloud. Trinity standardizes capabilities across sizes and is available on the Arcee API, OpenRouter, and Hugging Face.
Trinity Large is a 400-billion-parameter mixture-of-experts foundation model, among the largest open source models trained by a U.S. company. Trinity-Large-Thinking is a 399-billion-parameter text-only reasoning model released on April 1, 2026.