Trinity Large Thinking

Trinity Large Thinking is a reasoning-optimized 398B MoE agent model from Arcee AI, designed for extended thinking in long-horizon and multi-turn tool use. It is the first version of the trinity line, released on 2026-04-01. It accepts and returns text, with a context window of 262144 tokens and a maximum output of 262144 tokens. Weights are open. It supports reasoning mode and tool calling. Pricing starts at $0.25 per million input tokens and $0.80 per million output tokens, with cached input at $0.06 per million tokens. It is served by 5 hosts.

First version of its line in the registry.

• ·398B MoE architecture for reasoning-optimized tasks
• ·Extended thinking for long-horizon and multi-turn tool use
• ·262k token context and output windows
• ·Open weights with reasoning and tool calling modes

• Reach for it for long-horizon agentic tasks requiring extended reasoning
• Reach for it for multi-turn tool use with 262k token context
• Reach for it for complex text reasoning with open weights

5 hosts serve this model at the price above · the maker's documentation

The registry keeps the six most recent versions of a line. Older ones stay in the file and can be surfaced without a rebuild.

What is the context window and maximum output?

Both context window and maximum output are 262144 tokens.

Does it support tool calling and reasoning?

Yes, it supports both reasoning mode and tool calling.

Are the weights open and what is the price?

Weights are open. Pricing starts at $0.25 per million input tokens and $0.80 per million output tokens, with cached input at $0.06 per million.