Trinity-Large-Thinking
Open Source
About
A frontier open reasoning model for complex, long-horizon agents and multi-turn tool calling. Uses thinking tokens prior to responding, enabling stronger multi-turn tool use, better context coherence, cleaner instruction following, and more stable behavior across long-running agent loops.
Capabilities
VisionMultimodalReasoningFunction CallingTool UseJSON ModeCode Execution
Specifications
FamilyArcee Ai
Released2026-04-01
ArchitectureSparse Mixture of Experts (MoE)
SpecializationAgent reasoning
TrainingPost-trained