ChipAgents and NVIDIA launch Build-A-Chip-Bench: top AI model scores just 16.6% on end-to-end chip design
WilliamWangNLP · x · 2026-10-15
After two years in AI-native chip design, William Wang's ChipAgents partnered with NVIDIA to launch Build-A-Chip-Bench, the first benchmark testing whether AI can take a chip from specification through physical implementation to verified GDS.
Key findings:
- The top model scored just 16.6%; GPT-6 Astra 8.7%, Gemini 3.8 Flash 0.8%, Kimi K3 0.2%, GLM-5.3 zero
- Models can produce functionally correct RTL (Claude Fable 5.1: 65% functional pass, GPT-6 Astra: 85%), but reliably meeting timing, power, area, and physical-verification requirements remains a major gap
- Unlike RTL-only benchmarks such as CVDP and ChipBench, it jointly evaluates functional, GDS, and PPA-vs-constraint results
The takeaway: autonomous chip design is far harder than software engineering, and the benchmark sets a measurable target for progress.
Related event: NVIDIA and ChipAgents Launch Chip Design Benchmark; Best Agent Scores 16.6%(3 posts)→
More from coding & agent
- Sentry CEO says AI agent is now the company's No.1 code contributor — zeeg · 2026-10-15
- Dev: I accept about half of AI coding suggestions, a third surprise me positively — nickbaumann_ · 2026-10-15
- Armin Ronacher on why he prefers React after building with Solid.js, plus what's new with Pi — ianand · 2026-10-15
- Agents write most new code now — has your review process scaled? — ronbodkin · 2026-10-15
- 8k context beats 480k: OpenAI Decisions API cuts costs 84% at 98.25% F1 — OpenAIDevs · 2026-10-15
- Sentry CEO: AI agents still write messy code but are getting better at managing it — zeeg · 2026-10-15