Arcee AI Releases Trinity Large Thinking: An Apache 2.0 Open Reasoning Model for Long-Horizon Agents and Tool Use

Arcee AI released Trinity Large Thinking, an open-weight reasoning model in the Trinity Large family for agentic workflows.

The model uses a sparse MoE design with 400B total parameters, about 13B active per token, and a 262,144-token context window.

It is tuned for multi-turn tool calling and structured outputs, making it suited for long-context autonomous agents.

Trinity Large Thinking ranks #2 on PinchBench, behind Claude Opus.