AtelierEval - Prompt Constraint Coverage (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, mean of the open-ended, constrained and imitation categories, 120 tasks each; structured prompting with chain-of-thought and self-verification). Source: arxiv.org. Saturation forecast: Around May 2028. 8 models tracked.

Top models

#ModelScore
1Gemini 3 Pro (Preview)65.5
2GPT-5.263.7
3Claude Sonnet 4.560.1
4Qwen 3 VL 235B A22B58.8
5GPT-4.156.1
6GPT-4.1 Nano53.2
7Qwen 3 VL 8B52.7
8Gemini 2.0 Flash48.2

Interactive version: theaggregate.ai/benchmark?slug=ateliereval-prompt-constraint-coverage-skilled-prompting · How It Works · Data refreshed daily, snapshot 2026-09-26.