← AI news, last 7 days

OpenAI models for chip design

Last scan Sat Nov 21, 06:54 UTC
3 sources · 224 points on Hacker News · single origin (OpenAI)

OpenAI says its frontier models are improving chip design, and it reports up to 14% improvement over heavily optimized baselines on the flagship GPU IP in a partnership with AMD.12 According to OpenAI, the model rewrote SystemVerilog RTL and used Synopsys tools to check that behavior was preserved and to weigh power, performance and area; AMD has put some changes into a design sent for manufacturing and others into its production design files.3456 OpenAI adds that AMD confirmed improvements of a few percent in power, performance and area across major sections of its forthcoming top-tier CPU, based on designs the model produced.7

One multiplier redesign reportedly used about 15% less cell area in synthesis while cutting latency from two clock cycles to one.8

OpenAI also tested successive models on two open-source RISC-V cores, PicoRV32 and Ibex, and reports that GPT-6.2 Astra had the highest average score on both at a comparable budget of about 112,000 tokens.9 The full evaluation had 9,881 scored runs, and at about 112,000 tokens all five models scored higher when designing a core from written requirements than when improving the existing design.1011 OpenAI itself cautions that designing from a specification lets models omit features such as performance counters, which may explain part of the higher scores, and says requiring those features, tighter area limits and post-synthesis timing checks would make the test more representative of production work.1213

OpenAI says it is partnering with Synopsys to train a new model, GPT-Synopsys, to use Synopsys EDA tools the way expert chip designers do; a Latent Space newsletter relayed the AMD result as OpenAI's own report, and no independent assessment was reported. inference142

Every sentence was checked word for word against quotes from the articles it cites. Each footnote shows its quote, unless the page already quotes as much of that article as we allow; then it links the article instead. A model wrote and arranged the sentences, and a second read checked each against its quotes (written 2026-11-20).

Sources quoted

  1. We have partnered with AMD on PPA improvements and seen up to 14% improvement relative to heavily optimized baselines on the flagship GPU IP. — OpenAI, Advancing chip design with OpenAI models
  2. OpenAI reports up to 14% improvement against optimized baselines on blocks of AMD’s flagship GPU design. — Latent Space, [AINews] Gemini 4 Neon: GDM's Opus/Sol Contender
  3. It used Synopsys tools to verify that each rewrite preserved the block’s behavior and to evaluate the trade-offs in power, performance, and area. — OpenAI, Advancing chip design with OpenAI models
  4. AMD has already included some of these changes in a recent design it sent for manufacturing, and added others to its production design files for use in future versions. — OpenAI, Advancing chip design with OpenAI models
  5. The team used Synopsys tools to check that rewritten circuits preserved behavior and to measure power, performance and area. — Latent Space, [AINews] Gemini 4 Neon: GDM's Opus/Sol Contender
  6. Some changes made it into a design sent for manufacturing. — Latent Space, [AINews] Gemini 4 Neon: GDM's Opus/Sol Contender
  7. AMD also verified single-digit percentage gains in PPA across key blocks of its upcoming next-generation flagship CPU using designs written by our model. — OpenAI, Advancing chip design with OpenAI models
  8. One multiplier redesign used about 15% less cell area in synthesis while reducing latency from two clock cycles to one. — OpenAI, Advancing chip design with OpenAI models
  9. With a comparable budget of about 112,000 tokens per run, GPT‑6.2 Astra had the highest average score on both cores. — OpenAI, Advancing chip design with OpenAI models
  10. Quote not reprinted: this page already repeats as much of this article as we allow. — OpenAI, Advancing chip design with OpenAI models
  11. Quote not reprinted: this page already repeats as much of this article as we allow. — OpenAI, Advancing chip design with OpenAI models
  12. Quote not reprinted: this page already repeats as much of this article as we allow. — OpenAI, Advancing chip design with OpenAI models
  13. Quote not reprinted: this page already repeats as much of this article as we allow. — OpenAI, Advancing chip design with OpenAI models
  14. Quote not reprinted: this page already repeats as much of this article as we allow. — OpenAI, Advancing chip design with OpenAI models

Who reported it

Facts and takes

No checked facts or takes on this story yet.

Timeline

Built 2026-11-21 06:54 UTC. Facts and quotes say who states them, not that they are true.