Arcee, a tiny 26-person U.S. startup that constructed a large, 400B-parameter open source LLM on a $20 million shoestring price range, has launched its new reasoning model. Arcee calls the model Trinity Large Thinking — and it’s essentially the most succesful open-weight model “ever released by a non-Chinese company,” claims CEO Mark McQuade to TechCrunch.
As that remark implies, Arcee has a purpose that I can’t help however root for: It needs to present U.S. and Western corporations a model that offers them no cause to make use of a Chinese-based one.
While Chinese fashions are extraordinarily succesful, they’re perceived as dangerous, placing energy, and maybe information, into the arms of a authorities that doesn’t share all the Western world’s beliefs.
With Arcee, corporations can obtain the model, practice it to their very own wants, and apply it to premises. Companies also can use Arcee’s cloud-hosted model, accessible through API.
While Arcee’s fashions aren’t outperforming the closed source fashions from the large labs like Anthropic or OpenAI, they’re not being held hostage by the whims of these giants, both.
For occasion, Claude, with its distinctive skills to code, has been a preferred alternative for customers of open source AI agent software OpenClaw. But Anthropic pulled the rug out from them final week when it advised customers that their Anthropic subscriptions will not cowl OpenClaw utilization — they must pay moreover for that. (In February, OpenClaw creator Peter Steinberger stated he was becoming a member of Anthropic’s greatest rival, OpenAI.)
In distinction, McQuade proudly factors to information from OpenRouter that claims it has develop into one of many high fashions used with OpenClaw.
Techcrunch occasion
San Francisco, CA
|
October 13-15, 2026
So, how good is Trinity Large Thinking? It is similar to a number of the different high open source fashions, in response to the benchmark outcomes it shared with TechCrunch.

As we beforehand reported, it’s not a head-to-head menace to the large cheese amongst U.S.-built open fashions: Meta’s Llama 4. But it additionally doesn’t have the odd, not-really open source license problems with Meta’s model. All of Arcee’s Trinity fashions are launched below the gold customary for OS licenses, Apache 2.0.
Just to be clear, there are additionally numerous different U.S. startups providing open source fashions and, as a fan of the ingenuity of startups, I’m rooting for them, too.
