Released Trinity-Large-Thinking model as an open frontier model with reinforcement learning infrastructure support.