GDPval
Deliverables from 44 occupations judged blind against expert human work. Score is win rate.
Agent Index weight
7%
in the Knowledge-work agents pillar
Models tested
11
from the benchmark’s leaderboard and reports
Results
| # | Model | Score | Normalised | Harness | Cost/task | Date |
|---|---|---|---|---|---|---|
| 1 | GPT-5.2 Supported OpenAI | 49.7% | source | — | — | |
| 2 | Claude Opus 4.5 (latest) Supported Anthropic | 45.5% | source | — | — | |
| 3 | Claude Opus 4.1 (latest) Supported Anthropic | 43.6% | source | — | — | |
| 4 | Claude Sonnet 4.5 (latest) Supported Anthropic | 42.5% | source | — | — | |
| 5 | Gemini 3 Pro Preview Supported Google | 40.3% | source | — | — | |
| 6 | GPT-5 Supported OpenAI | 34.8% | source | — | — | |
| 7 | o3 Supported OpenAI | 30.8% | source | — | — | |
| 8 | o4-mini Estimated OpenAI | 25.3% | source | — | — | |
| 9 | Gemini 2.5 Pro Supported Google | 23.3% | source | — | — | |
| 10 | Grok 4 Supported xAI | 21.1% | source | — | — | |
| 11 | GPT-4o (2024-08-06) Supported OpenAI | 9.9% | source | — | — |
For models without a GDPval result, the index estimates one from Epoch’s Capabilities Index using a logistic fit on 11 models (residual SD 0.34 on the logit scale). Estimates are labelled and widen the model’s range.