Arcee AITrinity

Trinity Large Thinking

Trinity Large Thinking is Arcee's reasoning-focused 398B mixture-of-experts checkpoint, with about 13B active parameters.

What the release provides

The card lists a 512K extended context and OpenMDW-1.1 licensing. It instructs applications to retain the model's reasoning content in conversation history for subsequent turns and tool loops.Trinity Large Thinking official model card

How to use it well

Practical guidance. Keep the serving model's response fields intact through a tool cycle. Review parser behavior before comparing agent performance. Measure total memory and tool-task success with the intended context; active parameters alone do not describe deployment size.

Specifications & access

Trinity Large Thinking specifications
DetailValue and conditions
AccessDownloadable
Model identifiersarcee-ai/Trinity-Large-Thinking