GDPx Revolutionizes AI Benchmarking with Realistic, Context-Rich Tasks Across 11 Domains

November 2, 2026
GDPx Revolutionizes AI Benchmarking with Realistic, Context-Rich Tasks Across 11 Domains
  • GDPval is a dated benchmark, while GDPx shifts to interactive, context-rich tasks in realistic environments to better measure frontier capabilities.

  • GDPx positions itself as the successor to GDPval, emphasizing tasks set in real-like environments such as a private-equity deal room for finance rather than spoon-fed prompts.

  • Compared with GDPval’s non-interactive tasks, GDPx uses live cyber environments, simulated labs for materials science, hardware design sessions, and other realistic settings to reflect real-world usage.

  • Safety classifiers influence rankings, with top models like Claude Opus 5.5 and GPT-6 Astra penalized on the main leaderboard due to cybersecurity and biology domain restrictions.

  • Scores shift when safety domains are excluded, highlighting how safety classifiers shape model standings in cyber, biology, and alignment.

  • Some models’ scores appear limited by safety classifiers, with Opus 5.5 and GPT-6 Astra cited as being sandbagged in cybersecurity, biology, and alignment tasks.

  • GDPx uses an eight-hour default task duration, pass@1 across three attempts with equal domain weights, and involved vetting over 30 RL environment partners and reviewing hundreds of trajectories for failure modes.

  • SemiAnalysis’ figures underpin the scores, costs, and task counts; the 91% and 32% comparisons are from the article’s analysis, not re-run tasks.

  • Partnerships span domains with Aria, Arcverex, Ergodex for chip design; Arithmetic for cybersecurity; Latch Bio for biology; Living Context for finance; Chakra for legal; Third Moment for quant trading; Extrapolate for RSI; Beautiful Machines for hardware design; ValenceForge and Emulated for materials science; Fleet for alignment; and Datacurve for software engineering.

  • GDPx collaborates with specialized RL environment startups across domains to support task realism, including chip design, cybersecurity, biology, finance, legal, quant trading, hardware design, materials science, alignment, and software engineering.

  • Key domain partners include Aria, Arcverex, Ergodex; Latch Bio; Living Context; Chakra; Third Moment; Extrapolate; Beautiful Machines; ValenceForge/Emulated; Fleet; and Datacurve.

  • Top domains feature chip design, finance, and materials science, with all 11 domains weighted equally and scores averaged across three attempts per task.

Summary based on 5 sources


Get a daily email with more Tech stories

More Stories