Atlas

Benchmarks

← All benchmarks

GDPx: Materials Science

Science · 2026-11-01

GDPx's Materials Science domain: 23 tasks from ValenceForge and Emulated on controlling simulated lab equipment, from electron microscopes to battery cyclers, and interpreting raw experimental data. Each partner's pass@1 over three attempts per task is averaged, partners weighted equally. A component of the GDPx score, which weights its 11 domains equally.

Top models (higher is better)

ModelScore
GPT-6 Astra21.6
Gemini 4 Carbon20.6
GPT-6.1 Sol19.4
GPT-6.1 Terra18.2
Claude Sonnet 5.516.7
Claude Opus 5.516.5
Muse Wonder Full10.9
Gemini 4 Argon10.2
Claude Fable 5.19.7
Claude Haiku 5.57.2
Muse Wonder5.7
Kimi K35.7
GPT-6 Luna5.2
Grok 4.74.4
GLM-5.32.8
MiniMax M3.1 Flash2.3
Muse Spark Open2.0
Loading Atlas data…