ResearchOpenAIConfirmed
Story 12 of 18 in the Wednesday 18 November editionOpenAI cuts AMD GPU PPA up to 14% and partners on GPT-Synopsys
Published on 18 November 2026 alongside silicon-proven 10% to 13% Jalapeno area savings, 9,881 RISC-V runs and a Synopsys EDA model collaboration.
OpenAI reported on 18 November 2026 that its models rewrote SystemVerilog RTL to deliver up to 14% power, performance and area gains on AMD flagship GPU IP in under five months (with changes already sent for manufacturing), verified silicon area cuts of 13% and 10% on its custom Jalapeno inference chip, and partnered with Synopsys to train GPT-Synopsys on commercial EDA tools.
Study details
- AMD production silicon gains
- Up to 14% PPA gain on flagship GPU IP (sent to fab) plus single-digit PPA gains on next-gen flagship CPU
- Jalapeno inference chip cuts
- Silicon-proven 13% vector-unit and 10% matrix-unit standard-cell area reductions
- Open-source RISC-V benchmark
- 9,881 simulation-checked runs on 32-bit PicoRV32 and OpenTitan Ibex led by GPT-6.2 Astra (~112k tokens/run)
Release details & context
Why specification-first RISC-V designs beat RTL edits at 112k tokens
Across OpenAI’s 9,881 Yosys-scored evaluations within a 16-times area cap, GPT-6.2 Astra roughly doubled GPT-6 Astra on PicoRV32 RTL optimisation, and all five tested models scored higher when designing PicoRV32 and Ibex from written specifications (reaching roughly 0.7 and 0.75 mean throughput) than when editing existing RTL.
Sources
- PrimaryOpenAI: Advancing chip design with OpenAI modelsopenai.com
- On XOpenAI (@OpenAI) chip design announcement on Xx.com
OpenAI release timeline (51 launches tracked)
- OpenAI launches science blog with GPT-6 Astra Standard Model proof
OpenAI debuts its science blog with UCSB physicist Nathaniel Craig’s candidate nonperturbative Standard Model lattice proof built with GPT-6 Astra.
- OpenAI expands ChatGPT for Academic Researchers to 200,000 scientists
OpenAI opens 50,000 ChatGPT for Academic Researchers grants on 23 November for up to 200,000 scientists, reporting 1.9 times faster work.
- OpenAI launches GPT-6.2 Astra and GPT-6.2 Sol at $8 and $2 input pricing
OpenAI ships GPT-6.2 Astra ($8/$40 per million tokens, down 20%) and GPT-6.2 Sol ($2/$10) across the API, Codex, ChatGPT Work and partner platforms.
- OpenAI introduces ChatGPT for Sales with 30+ CRM and revenue plugins
OpenAI launches ChatGPT for Sales, combining a native Sales plugin, persistent dot agents and more than 30 CRM, call, prospecting and CPQ connectors.

