AtelierEval - Prompt Constraint Coverage - Imitation (Skilled Prompting): leaderboard
Metric: Prompt checklist coverage (%; share of each task's expert constraint checklist that the written prompt specifies, checked by a GPT-5.4 AtelierJudge; 120 imitation tasks: write a prompt that reproduces a given target image; structured prompting with chain-of-thought and self-verification). Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 66.6 |
| 2 | GPT-5.2 | 64.6 |
| 3 | Claude Sonnet 4.5 | 59 |
| 4 | Qwen 3 VL 235B A22B | 52.8 |
| 5 | Qwen 3 VL 8B | 46.3 |
| 6 | GPT-4.1 | 45.1 |
| 7 | GPT-4.1 Nano | 43.4 |
| 8 | Gemini 2.0 Flash | 34.4 |
Interactive version: theaggregate.ai/benchmark?slug=ateliereval-prompt-constraint-coverage-imitation-skilled-prompting · How It Works · Data refreshed daily, snapshot 2026-09-26.