NVIDIA says Nemotron 3 Ultra leads open models in agentic RTL design
imjustnewatai · x · 2026-07-27
NVIDIA says Nemotron 3 Ultra beats GLM 5.2 and Kimi K2.6 on agentic RTL design
NVIDIA’s technical blog says Nemotron 3 Ultra powers an agentic chip-design workflow for RTL generation and verification. In the same ACE-RTL pipeline, it outperforms GLM 5.2 and Kimi K2.6 across nine CVPD task categories.
Key claims from the post:
- 97.1% average pass rate across nine hardware tasks
- 100% on several categories, including debugging
- 71% fewer tokens per iteration than Kimi K2.6
- Synopsys says its Nemotron-powered agent can run the full chip verification cycle and reach validated RTL up to 50x faster with 20% more coverage
- Siemens says its self-verifying agents can reach signoff-quality results in hours instead of days
The post frames this as a feedback loop: AI improves chip design, better chips strengthen AI, and the cycle repeats. It also links the theme to Anthropic’s hiring for model work on silicon design.
Related event: NVIDIA Claims Nemotron 3 Ultra Hits 97.1% in RTL Chip Design(3 posts)→
More from coding & agent
- Annotate lets AI agents draw arrows and labels on your screen via MCP — adamm_carter · 2026-07-27
- Moonshot says Kimi K3 open weights are coming, and asks whether developers really use Chinese models — SwordfishGreedy1945 · 2026-07-27
- Open-source agent architecture aims to keep cache hits while controlling context growth — dddhhyyw · 2026-07-27
- Weaker models expose harness bugs that frontier models quietly work around — arsenyinfo · 2026-07-27
- A new @asciidotdev agent set up a sandbox, synced files and found the available CLIs — jasonkneen · 2026-07-27
- Claude Code’s ultracode mode powers a browser game build, but burns $2,500 in tokens — majidmanzarpour · 2026-07-27