GDPx
Professional Work · 2026-11-01
SemiAnalysis's GDPx v1.0 runs 206 agentic tasks built by 14 domain-specialist partners across 11 domains: chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement, hardware design, materials science, alignment and software engineering. Each model works in its native agent harness (OpenCode when it has none) inside a sandboxed environment, with an eight-hour default limit per task. Tasks are graded binary or with partial credit depending on the domain, and a task scores pass@1 over three attempts. The GDPx score is the mean of the 11 domain scores, each the mean of its partners.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 12.0 |
| Gemini 4 Carbon | 10.9 |
| GPT-6 Astra | 9.6 |
| Muse Wonder Full | 9.0 |
| Claude Fable 5.1 | 8.5 |
| GPT-6.1 Terra | 8.5 |
| GPT-6.1 Sol | 8.4 |
| Claude Sonnet 5.5 | 8.2 |
| Gemini 4 Argon | 7.3 |
| Muse Wonder | 6.8 |
| GPT-6 Luna | 5.1 |
| Claude Haiku 5.5 | 4.9 |
| MiniMax M3.1 Flash | 4.0 |
| Grok 4.7 | 3.9 |
| Kimi K3 | 3.7 |
| GLM-5.3 | 2.7 |
| Muse Spark Open | 1.7 |