AI for electrical engineering and chips, right now
Designing circuits and boards, writing hardware code, chip design and verification.
AI models now assist chip and PCB design, with Claude Fable 5.2 shown to design working electronic prototypes end-to-end.
What models can do now
- Claude Fable 5.2 designed three functional electronic devices, including schematics, PCB layouts, firmware, and parts lists, with no circuit corrections needed.[5]
- Magic Gourd, an interactive AI agent, generates synthesizable RTL code, testbenches, and documentation, achieving 15-25x productivity improvements in chip front-end design.[4]
- ChipOmni, a general-purpose AI agent, covers over 50 EDA tasks including netlist, RTL, and testbench generation, and formal verification.[3]
- OpenAI reports its models help engineers design, simulate, and build physical products, including electrical engineering, and tested GPT-6 Astra in multimodal workflows.[6]
- Electrode 1.0, an open-weight language model for KiCad, performs schematic and board operations using 60+ tool calls, scoring 0.5388 on a 100-task test.[7]
Where they still fall short
- No benchmark directly measures AI performance in electrical engineering; the closest general benchmarks are Claude Sonnet 5.5, Gemini 4 Argon, and GPT-6 Astra.[1]
- Earlier Claude iterations struggled with component placement and routing connections on PCBs, often producing incomplete or impractical designs.[5]
How fast it is moving
The record on this field's benchmarks over the last year, from Spezia's model leaderboard.
No public benchmark measures electrical engineering directly yet. The closest are hands-on technical work in a terminal, scientific coding and graduate physics.
Hard jobs done in a real computer terminal: building and fixing software, setting up systems, and science and engineering problems.
Record: 70.6%, up from 55.8% 1 months ago (Claude Sonnet 5.5)
Research coding problems written by scientists across 16 fields, from physics and chemistry to biology and materials science.
Record: 91.8%, up from 44.8% a year ago (2×) (Mistral Large 4)
- GPQA Diamond96.5%, up from 87.6% a year ago · Gemini 4 Carbon
The best models for electrical engineering now
On the closest general benchmarks, since none measures this field directly.
- 1
Claude Sonnet 5.5
- 2
Gemini 4 Argon
- 3
GPT-6 Astra
- 4
Claude Fable 5.2
- 5
Kimi K3.1 Pro
- 6
Gemini 3.1 Pro
Milestones
What became possible, newest first, each marked by how firm it is. Open one for the words it rests on.
Showndemonstrated with published results, a paper or an independent checkClaimedsaid by the company or people who built it, not yet checked by othersReportedreported by others, not confirmed by the company or a published resultNot publicdone by a model that is not released to the public
- Nov 9, 2026ShownClaude Fable 5.2
Claude Fable 5.2 achieved end-to-end electronic prototype design, producing functional devices without circuit corrections.
“Anthropic's Claude Fable 5.2 demonstrates the ability to design and prototype electronic devices, showcasing advancements in geometric reasoning and PCB layout.” Claude Fable 5.2 Achieves End-to-End Electronic Prototype Design
- Nov 8, 2026ShownChipOmni
ChipOmni, a general-purpose AI agent for EDA, was introduced, covering over 50 tasks with a knowledge base of 10,000+ documents.
“Its knowledge base contains more than 10,000 textbooks and technical documents and over 20 EDA software manuals.” ChipOmni: A General-Purpose AI Agent for Electronic Design Automation
- Oct 30, 2026ShownMagic Gourd
Magic Gourd, an interactive AI agent for chip front-end design, was published, demonstrating 15-25x productivity improvements.
“Experimental results demonstrate 15-25\times productivity improvements in design tasks.” Magic Gourd: An AI Agent for Interactive Chip Front-End Design
- Oct 22, 2026ShownPCBgen
PCBgen, a self-evolving LLM agent for specification-to-schematic PCB design with parameter tuning, was published at DAC.
“PCBgen: Self-Evolving LLM Agent for Specification-to-Schematic PCB Design with Parameter Tuning” PCBgen: Self-Evolving LLM Agent for Specification-to-Schematic PCB Design with P
- Oct 18, 2026ShownElectrode 1.0
Electrode 1.0, a domain-adapted open-weight language model for KiCad PCB design automation, was introduced, scoring 0.5388 on a 100-task test.
“Electrode 1.0 gets 0.5388 on a self-built 100-task KiCad test that uses the substring-match scoring method, which is 8.4 % ahead of baseline Qwen-Coder model” Domain-adapted open-weight language model for KiCad printed-circuit-board design
Sources
- [1]Best models on the general benchmarks closest to electrical engineering (none measures it directly)Leaderboard · 2026-11-12
- [2]Beyond Automation: The Emerging Role of AI in RF and Microwave Design - MTT-SWeb · 2026-11-01
- [3]ChipOmni: A General-Purpose AI Agent for Electronic Design AutomationWeb · 2026-11-08
- [4]Magic Gourd: An AI Agent for Interactive Chip Front-End DesignWeb · 2026-10-29
- [5]Claude Fable 5.2 Achieves End-to-End Electronic Prototype DesignWeb · 2026-11-09
- [6]Mempercepat rekayasa untuk dunia fisik - Hari ini, kami membagikan kemajuan dalam cara model OpenAI membantu insinyur merancang, menyimulasikan, dan membuat produk fisikWeb · 2026-11-02
- [7]Domain-adapted open-weight language model for KiCad printed-circuit-board design automationWeb · 2026-10-18
Other fields
- AI for math
- AI for biology and drug discovery
- AI for medicine and health
- AI for physics, chemistry and materials
- AI for software engineering
- AI for game development
- AI for content and media production
- AI for finance
- AI for law
- AI for remote and knowledge work
- AI for cybersecurity
- AI for robotics and the physical world
- AI for research and analysis