Arcee AI Launches Trinity Large Thinking: New Open Reasoning Model

The landscape of open-source artificial intelligence has shifted from purely generative models toward systems capable of complex, multi-step reasoning. While proprietary ‘reasoning’ models have dominated the conversation, Arcee AI has released Trinity Large Thinking. This release is an open-weight reasoning model distributed under the Apache 2.0 license, positioning it as a transparent alternative for developers building autonomous agents.

Unlike models optimized solely for conversational chat, Trinity Large Thinking is specifically developed for long-horizon agents, multi-turn tool calling, and maintaining context coherence over extended workflows. The architecture of Trinity Large Thinking is based on a sparse Mixture-of-Experts (MoE) model with 400 billion parameters, but it activates only 13 billion parameters per token using a 4-of-256 expert routing strategy for inference efficiency. This sparsity provides the world-knowledge density of a massive model without the prohibitive latency typical of dense 400B architectures.

Key technical innovations in the Trinity Large family include a new MoE load balancing strategy called SMEBU, which prevents expert collapse and ensures more uniform utilization of the model’s specialized pathways. Arcee utilized the Muon optimizer during the training of the 17-trillion-token pre-training phase, allowing for higher capital and sample efficiency compared to standard AdamW implementations. The model also features an attention mechanism that interleaves local and global attention alongside gated attention to enhance its ability to comprehend and recall details within large contexts.

A core differentiator of Trinity Large Thinking is its behavior during the inference phase. The Arcee team states that the model utilizes a ‘thinking’ process prior to delivering its final response. This internal reasoning allows the model to plan multi-step tasks and verify its logic before generating an answer. Trinity Large Thinking is optimized for the ‘Agentic’ era, and its performance is measured not just on general-knowledge trivia but on its reliability in complex software environments.

The model has demonstrated strong performance in PinchBench, a benchmark designed to evaluate model capability in environments relevant to autonomous agents. Currently, Trinity Large Thinking holds the #2 spot on PinchBench, trailing only behind Claude Opus-4.6. The model supports a context window of 262,144 tokens, enabling it to process massive datasets or long conversational histories for agentic loops. Training focused heavily on multi-turn tool use and structured outputs, ensuring that the model can call APIs and extract parameters with high precision over many turns.