英伟达刚刚发布了Llama Nemotron Ultra。
NVIDIA刚刚发布了Llama 3.1 Nemotron Ultra（253亿参数模型），在GPQA-Diamond、AIME和LiveCodeBench上表现出色。<p>他们的博客详细介绍了这一点，显示出其吞吐量比DeepSeek-R1高出最多4倍，并且基准测试结果更佳。<p>博客链接：
https://developer.nvidia.com/blog/build-enterprise-ai-agents-with-advanced-open-nvidia-llama-nemotron-reasoning-models<p>该模型已在HF和NIM上提供。有人尝试过吗？<p>HF链接：
https://huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1<p>NIM链接：
https://build.nvidia.com/nvidia/llama-3_1-nemotron-ultra-253b-v1
查看原文
NVIDIA just released Llama 3.1 Nemotron Ultra (253B parameter model) that’s showing great performance on GPQA-Diamond, AIME, and LiveCodeBench.<p>Their blog goes into detail but it shows up to 4x throughput over DeepSeek-R1 with better benchmarks.<p>Blog:
https://developer.nvidia.com/blog/build-enterprise-ai-agents-with-advanced-open-nvidia-llama-nemotron-reasoning-models/<p>The model is available on HF and as a NIM. Has anyone tried it?<p>HF:
https://huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1<p>NIM:
https://build.nvidia.com/nvidia/llama-3_1-nemotron-ultra-253b-v1