新闻
NVIDIA Nemotron 3 Ultra Leads Open Models on Accuracy and Efficiency in Agentic RTL Coding
NVIDIA Developer · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- NVIDIA released Nemotron 3 Ultra, achieving 97.1% pass rate on RTL coding tasks, outperforming competitors while using 71% fewer tokens.
- 为何重要
- Hardware engineers and chip designers using AI agents for register transfer level code generation, modification, debugging, and verification workflows.
- 注意
- Results measured on CVDP benchmark with ACE-RTL agent framework; real-world performance on production designs and integration with EDA tools requires validation.
收听本摘要
文章节选
Agentic AI / Generative AI
NVIDIA Nemotron 3 Ultra Leads Open Models on Accuracy and Efficiency in Agentic RTL Coding
Jul 26, 2026
By Nirmal Kumar Juluru , Chenhui Deng , Cunxi Yu , Nathaniel Pinckney and Brucek Khailany
Like
Discuss (0)
L
T
F
R
E
AI-Generated Summary
Like
Dislike
The ACE-RTL agent, in combination with NVIDIA Nemotron 3 Ultra, delivers state-of-the-art accuracy and efficiency in agentic RTL workflows by leveraging an iterative generate-test-reflect approach and long-context, RTL-specialized reasoning.
On the comprehensive Verilog design problems (CVDP) benchmark, ACE-RTL with Nemotron 3 Ultra achieves a 97.1% average pass rate across nine agentic RTL task categories, outperforming models like GLM 5.2 and Kimi K2.6 while using up to 71% fewer tokens per iteration.
Nemotron 3 Ultras hybrid Mamba-Attention Mixture-of-Experts architecture and training on a diverse, rubric-filtered, synthetic RTL dataset enable high-throughput, low-cost, long-context inference, making it a practical foundation for integration with leading EDA tools from Cadence, Siemens, and Synopsys.
AI-generated content may summarize information incompletely. Verify important information. Learn more
Modern chip design is increasingly limited by engineering time. Register transfer level (RTL) development and verification require specialized hardware knowledge, precise reasoning, a
节选自原文。请前往来源阅读全文。
- agent
- agentic
- llm
- reasoning
- edge
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。