NVIDIA Launches OpenReasoning-Nemotron Reasoning Model

On July 20, NVIDIA launched OpenReasoning-Nemotron, a new suite of inference models that includes four Qwen-2.5 fine-tuned models with parameter sizes of 1.5B, 7B, 14B, and 32B, all derived from the 671-billion-parameter DeepSeek R1 0528 large model. Through the process of "distillation", NVIDIA has successfully compressed this mega-model into a lighter inference model, lowering the deployment threshold so that advanced inference experiments can be conducted even on standard gaming computers, avoiding the high cost of GPUs and cloud computing.