Intel / post

Arcee launches Trinity, a new open-weight MoE family

Source: x.com
Date: Published Dec 1, 2025

Context

Arcee introduced Trinity, its first model family, releasing two sizes at once on December 1, 2025: Trinity-Mini, with 26 billion total parameters but only 3 billion active per , and a smaller Trinity-Nano-Preview at 6 billion total and 1 billion active. Because only a fraction of each model's parameters run on any given token, both are cheaper to serve than a dense model of the same total size.

Both models are downloadable from Hugging Face, and Arcee made them free to use for a limited time through OpenRouter, an API marketplace that hosts many providers' models behind one interface, giving developers a way to try the models in real workloads before deciding whether to host them.

Terms in this piece · Glossary

open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.

Read the source x.com