Knowledge work
GDPval (win/tie rate)
How often the AI's work matches or beats a human expert's on real knowledge-work tasks. Higher is better.
Rankings
Higher is betterGDPval (win/tie rate) — frequently asked questions
- What is GDPval (win/tie rate)?
- How often the AI's work matches or beats a human expert's on real knowledge-work tasks. Higher is better.
- Which AI model scores highest on GDPval (win/tie rate)?
- GPT-5.5 by OpenAI holds the best GDPval (win/tie rate) result among tracked models, at 84.9% (released Apr 23 2026). Higher scores are better on this benchmark.
- What are the top 5 models on GDPval (win/tie rate)?
- 1. GPT-5.5 (OpenAI) — 84.9%; 2. GPT-5.4 (OpenAI) — 83%; 3. GPT-5.5-Pro (OpenAI) — 82.3%; 4. GPT-5.4-Pro (OpenAI) — 82%; 5. Claude Opus 4.7 (Anthropic) — 80.3%.
- How many models have a published GDPval (win/tie rate) score?
- 6 tracked models have a published GDPval (win/tie rate) score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.