inclusionAI/Ring-flash-2.0

Trained on over 20 trillion tokens and refined with supervised fine-tuning and multi-stage reinforcement learning, the model demonstrates strong performance against dense models up to 40B parameters. It excels in complex reasoning, code generation, and frontend development.