GDPval (OpenAI Evals): leaderboard
220 real-world tasks from 44 occupations across the top 9 GDP sectors. Built with industry professionals averaging 14 years of experience. Evaluates correctness, structure, style, and relevance.
Metric: Win Rate (%). Source: evals.openai.com. Saturation forecast: Around January 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (High) | 49.88 | #110 (GPT-5.2) |
| 2 | Claude Opus 4.5 | 45.37 | #83 |
| 3 | Claude Opus 4.1 | 43.77 | #140 |
| 4 | Claude Sonnet 4.5 | 42.63 | #142 |
| 5 | Gemini 3 Pro | 40.25 | #80 |
| 6 | GPT-5 (High) | 34.77 | #98 (GPT-5) |
| 7 | O3 (High) | 30.75 | #124 (O3) |
| 8 | O4 Mini (High) | 25.15 | #180 (O4 Mini) |
| 9 | Gemini 2.5 Pro | 23.42 | #152 |
| 10 | Grok 4 | 21.23 | #181 |
| 11 | GPT-5 (Medium) | 14.96 | #98 (GPT-5) |
| 12 | GPT-5 (Low) | 14.02 | #98 (GPT-5) |
| 13 | O3 (Medium) | 12.53 | #124 (O3) |
| 14 | O3 (Low) | 11.8 | #124 (O3) |
| 15 | GPT-4o | 9.85 | #343 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=gdpval-openai-evals · How It Works · Data refreshed daily, snapshot 2026-10-20.