Atlas

Benchmarks

← All benchmarks

GDPx: Hardware Design

Professional Work · 2026-11-01

GDPx's Hardware Design domain: 21 tasks from Beautiful Machines on CAD, CAM, PCB layout and FPGA work in real engineering software, seeded from hardware that actually shipped. Scores are pass@1 over three attempts per task. A component of the GDPx score, which weights its 11 domains equally.

Top models (higher is better)

ModelScore
Claude Opus 5.518.7
GPT-6 Astra16.5
GPT-6.1 Sol11.4
Claude Sonnet 5.510.0
GPT-6.1 Terra9.3
Claude Fable 5.18.3
Gemini 4 Carbon5.8
Claude Haiku 5.53.1
Gemini 4 Argon2.7
Kimi K32.3
MiniMax M3.1 Flash2.2
GPT-6 Luna2.0
Muse Wonder Full1.6
Muse Wonder1.5
GLM-5.31.2
Grok 4.70.5
Muse Spark Open0.2
Loading Atlas data…