Arcee releases 400B-parameter open-weight reasoning model

Arcee, a 26-person U.S. startup, has launched a new reasoning model—a type of artificial intelligence designed for complex problem solving—called Trinity Large Thinking. The company built the 400B-parameter open-weight model—an AI model where the weights are made publicly available—on a $20 million budget. By offering an open-weight architecture, Arcee allows organizations to download the model directly, train it to their specific operational needs, and run it on-premises. Alternatively, companies can access the model through Arcee’s cloud-hosted version via an API.

Mark McQuade, the CEO of Arcee, claims that Trinity Large Thinking is the most capable open-weight model “ever released by a non-Chinese company.” The startup is positioning the release as a direct alternative to Chinese-developed models. According to the source, Chinese models are perceived as risky, putting power, and perhaps data, into the hands of a government that doesn’t share all of the Western world’s ideals. To address these concerns for Western enterprises, Arcee has released its model under the Apache 2.0 license. This licensing choice distinguishes it from other U.S.-built open models, such as Meta’s Llama 4, which carry more restrictive licensing terms. While Trinity Large Thinking is not a head-to-head threat to Meta’s Llama 4, it avoids the licensing complexities associated with Meta’s model.

While Arcee’s models are not outperforming closed-source models from major artificial intelligence labs like Anthropic or OpenAI, they provide an alternative for developers seeking to avoid restrictive ecosystem changes. For example, users of OpenClaw, an open-source AI agent tool, faced disruptions last week when Anthropic informed users that their subscriptions would no longer cover OpenClaw usage, requiring additional payments. This shift followed an announcement in February when OpenClaw creator Peter Steinberger stated he was joining OpenAI, Anthropic’s primary competitor. By offering a model under the Apache 2.0 license, Arcee provides a stable option for developers who want to avoid being subject to the shifting subscription policies of larger AI labs.

Why it matters

Arcee is positioning its new model as a strategic alternative for Western organizations wary of relying on Chinese-developed AI, emphasizing open-weight accessibility over closed-source dependency.