Llama-3.3 Nemotron Super 49B

Llama-3.3-Nemotron-Super-49B-v1 is a 49-billion-parameter model from NVIDIA, fine-tuned from Meta's Llama 3.3 using NVIDIA's Nemotron post-training pipeline that combines supervised fine-tuning with reinforcement learning to enhance reasoning, instruction alignment, and complex problem-solving. The Super tier in the Nemotron family represents a mid-range capability level — positioned above the Nano series and below the Ultra 253B flagship — offering a balance between high-quality outputs and manageable inference infrastructure requirements. Released open-weight on HuggingFace with NVIDIA NIM support, it targets teams with multi-GPU setups who need strong reasoning capability without the scale of the Ultra model.