SemiAnalysis launches GDPx, an 11-domain agentic benchmark
The best model scores 12%, Claude Opus 5.5 leads overall
Confirmed
The takeaway
On Nov 2, 2026 SemiAnalysis published GDPx, a benchmark of long agentic tasks in realistic environments (deal rooms, sandboxed engineering software, live cyber ranges) across 11 economically important domains.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 2 November 2026
Your AI and this story
- GPT-6 Astra186 days after its cutoff
- Claude Opus 5.5125 days after its cutoff
- Gemini 4 Argon216 days after its cutoff
- Grok 4.7155 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 125 days before it.
Key facts
- Domains: chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement (RSI), hardware design, materials science, alignment, software engineering
- Scoring: pass@1 over three attempts, domains weighted equally; each domain built with at least one specialised data partner, SemiAnalysis did final QA
- Examples (SemiAnalysis): chip-design task vt01, GPT-6 Astra 0.34 vs Claude Opus 0.135 vs Fable 0.07; legal redline task, Opus 5.5 44% partial score, GPT-6 Astra 25%, Gemini 4 Argon 2%, GLM-5.3 0.5%, binary 0% for all
- Finance and legal: nearly every model scores 0 under binary grading
What happened
See summary. Overall per-model scores are shown only in charts on the page; we quote only numbers stated in the text.
Why it matters
A new, hard, multi-domain agentic benchmark from an independent analyst firm, with an RSI domain, published while labs’ own RSI claims are contested. It is new and not yet replicated.
Sources
2 sources from 2 sites. Numbers match the chips in the text.
2 sources: 1 primary, 1 reaction
Primary
- SemiAnalysis: GDPx, measuring frontier model capabilities (Nov 2)newsletter.semianalysis.com, official
Reactions
- SemiAnalysis on X (thread)x.com, discussion
Changes
- Filed