NVIDIA测了Nemotron 3 Ultra在芯片设计RTL编码上的表现,九类任务平均通过率97.1%,开源模型里最强,值得关注。
NVIDIA在代理芯片设计场景中测试了Nemotron 3 Ultra的RTL编码能力。该模型迭代编写芯片逻辑代码、通过仿真器运行、读取失败并重写。在九个真实设计工作类别中,Nemotron 3 Ultra平均通过率达97.1%,每次迭代消耗6629个tokens。这是测试中所有开源模型里表现最好的。
We tested Nemotron 3 Ultra on agentic chip design. The task is RTL coding: a model iteratively wri...
We tested Nemotron 3 Ultra on agentic chip design. The task is RTL coding: a model iteratively writes the code that defines a chip's logic, runs it through a simulator, reads the failures, and rewrites. Across nine categories of real design work, Nemotron 3 Ultra averaged a 97.1% pass rate at 6,629 tokens per iteration, best of the open models tested on both. 💬 5 🔄 11 ❤️ 52 👀 3931 📊 12 ⚡