Is GPQA Diamond still meaningful?

Known problems

• Epoch AI sets the ceiling at 97%, so the top models sit in the noise.^{[1][1] Benchmark data bundle Epoch AI Primary}
• It went from 50% to 90% in 17 months; the last points near the ceiling come slowly.^{[1][1] Benchmark data bundle Epoch AI Primary}
• Anthropic dropped it from its Claude Opus 5.5 launch table (22 Sep 2026); OpenAI still reports it.^{[2][2] Introducing Claude Opus 5.5 Anthropic Vendor[3][3] GPT-6 Astra benchmarks explained Vellum Secondary}

What to look at instead

• Humanity's Last Exam (HLE) Status:Flawed Use with care. Nearly half of a sampled set was defective, so top scores partly reflect broken items.
• HLE-Diamond Status:New / unproven Promising but unproven. It launched in September 2026 and has no independent audit yet.

Where it appears in the report

Chapter 1, Benchmarks

• Figure 1. Each New Benchmark Fills Up Faster Than the Last
• Figure 2. Recent benchmarks went from 50% to 90% in under a year
• Figure 3. How Fast Do Benchmarks Saturate?
• Table 2. First model to cross each threshold in “Go deeper: The full saturation timeline (14 benchmarks)”
• Table 5. Reasoning and knowledge benchmarks: what they measure, who maintains them, the top score on 5 November 2026 and the known flaw in “Go deeper: Reasoning and knowledge benchmarks (17)”
• Figure 14. OpenAI and Anthropic share only five headline benchmarks