Jason AI
Mini toolsTrinity Large Preview
New
Text
Gen 7
Overview
Trinity Large Preview is a 400B-parameter (about 13B active) frontier Trinity MoE model with long-context comprehension, tuned for strong reasoning, coding and multi-step agents, served via hosted preview APIs.
Description
Trinity Large Preview is Arcee’s largest open-weight Trinity model so far, a 400B sparse MoE with roughly 13B active parameters per token using 4-of-256 expert routing. Trained on more than 17T tokens, it targets frontier-level performance on reasoning-heavy workloads such as math, coding and complex multi-tool agent workflows, while staying efficient enough for practical long-context inference. The Preview checkpoint is lightly post-trained from Trinity Large Base and is accessible through Hugging Face, OpenRouter and several gateways.
Last updated: January 30, 2026
