AtelierEval - Prompt Constraint Coverage - Constrained Creation (Novice 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 constrained creation tasks: structured multi-constraint specifications; direct natural-language prompting). Source: arxiv.org. Saturation forecast: Around 2031. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 37.5 |
| 2 | Gemini 3 Pro (Preview) | 36.2 |
| 3 | GPT-4.1 | 32.9 |
| 4 | Qwen 3 VL 235B A22B | 31.9 |
| 5 | GPT-4.1 Nano | 28.4 |
| 6 | Claude Sonnet 4.5 | 27.8 |
| 7 | Qwen 3 VL 8B | 26.5 |
| 8 | Gemini 2.0 Flash | 22.8 |
Interactive version: theaggregate.ai/benchmark?slug=ateliereval-prompt-constraint-coverage-constrained-creation-novice-prompting · How It Works · Data refreshed daily, snapshot 2026-09-26.