Monday, November 2, 2026HotTea verified storyVerified 12:31 AM PST
← Back to the Monday, November 2, 2026 edition

SemiAnalysis's new AI work test tops out at 12 percent

Claude Opus 5.5 leads a 206-task benchmark. Its weighted score measures performance on selected tests, not a share of all jobs.

Verified 12:31 AM PST · 2 original sources

SemiAnalysis released GDPx on November 2. It has 206 tasks from 14 specialist data partners across 11 fields. Its published leaderboard puts Claude Opus 5.5 first at 12 percent overall.

Each task puts an agent in a working environment, including document collections, engineering software and simulated laboratories. SemiAnalysis reports single-attempt performance across three attempts, with equal weight for each field. The default task limit is eight hours, with shorter limits for some tasks.

SemiAnalysis and its partners designed and scored this test. These are their results, not an independent replication or a measure of the share of jobs AI can do. The chosen fields reflect the publisher's view of likely AI demand. SemiAnalysis also says safety filters lower some models' scores in cybersecurity, biology and alignment.

The test asks a model to find information, use tools and produce work that meets its requirements. A short-answer test does not check all of those steps.

Watch for outside replication, repeat tests on comparable tasks, and costs that include review and correction. Compare results in the relevant field before choosing a model.

Audit the story

Original sources

Company claims remain company claims. Follow the reporting and judge the evidence directly.

  1. SemiAnalysisGDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains ↗
  2. SemiAnalysisGDPx version 1.0 leaderboard and task overview ↗

Continue the edition

Read the full edition.

Read the full editionListen to the daily audio →