Model

Llama 3.1 Nemotron Ultra 253B v1

NVIDIA

开源 Llama 3.1 Community License

A 253B parameter derivative of Meta Llama 3.1 405B Instruct, developed by NVIDIA using Neural Architecture Search (NAS) and vertical compression. It underwent multi-phase post-training (SFT for Math, Code, Reasoning, Chat, Tool Calling; RL with GRPO) to enhance reasoning and instruction-following. Optimized for accuracy/efficiency tradeoff on NVIDIA GPUs. Supports 128k context.

发布日期2025年4月7日

参数规模253B

上下文长度—

许可证Llama 3.1 Community License

知识截止2023年12月1日

论文模型权重官方博客在线体验对比此模型

Benchmarks

评测成绩

评测基准	类别	分数	来源
MATH-500	math reasoning	97.0	来源
IFEval	general	89.5	来源
GPQA	reasoning general	76.0	来源
BFCL v2	general reasoning	74.1	来源
AIME 2025	math reasoning	72.5	来源
LiveCodeBench	reasoning general code	66.3	来源

Pricing

API 价格对比

暂无 API 价格。