GDPx
by SemiAnalysis (Max Kan, Andrew Megalaa, Dylan Patel)
What is GDPx? GDPx is an agents & tool use eval made by SemiAnalysis (Max Kan, Andrew Megalaa, Dylan Patel). A benchmark of 206 long, realistic agent tasks across 11 domains, including chip design, cybersecurity, biology, finance, legal, quant trading, hardware design, materials science, alignment and software engineering. Tasks come from 14 data partners that build reinforcement-learning environments. Each model runs in its own native coding harness with a default 8-hour limit, and the headline score is pass@1 over three attempts with every domain weighted equally. At launch on 2 Nov 2026 the top score was 12.0%, so the tasks are far from solved. The authors call it a successor to GDPval and plan a new version every 3 to 6 months.
Editor note
A benchmark of 206 long, realistic agent tasks across 11 domains, including chip design, cybersecurity, biology, finance, legal, quant trading, hardware design, materials science, alignment and software engineering. Tasks come from 14 data partners that build reinforcement-learning environments. Each model runs in its own native coding harness with a default 8-hour limit, and the headline score is pass@1 over three attempts with every domain weighted equally. At launch on 2 Nov 2026 the top score was 12.0%, so the tasks are far from solved. The authors call it a successor to GDPval and plan a new version every 3 to 6 months.
- What it tests
- Whether frontier agents can finish multi-hour expert work in real technical and professional environments.
- Original source
- gdpx.semianalysis.com
- Last updated
- Launched 2026-11-02
- License
- Not checked yet — see source
Related agents & tool use evals
- GAIA — General assistant skills: browsing, file handling, multi-step reasoning.
- WebArena — Autonomous web navigation and task completion.
- OSWorld — Computer-use agents operating real operating systems and apps.
- Terminal-Bench — Agents working in the shell end to end.
Discussion
No one has discussed GDPx yet. Have you run it? Share what it caught, what it missed or how you set it up.