Capabilities
VisionMultimodalReasoningFunction CallingTool UseStructured OutputsCode Execution
About Trinity Nano
6B sparse MoE with ~1B active parameters per token and 128K context window. Trained on 10T tokens. Personality-forward chat model designed for edge deployment and fast inference; experimental/preview status. Open-source under Apache 2.0.
Get Started
Model Specs
Released2025-02-01
Parameters6B
Context128K
ArchitectureSparse Mixture of Experts (MoE)