AtelierEval - Prompt Constraint Coverage (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, mean of the open-ended, constrained and imitation categories, 120 tasks each; direct natural-language prompting). Source: arxiv.org. Saturation forecast: Around September 2028. 8 models tracked.

Top models

#ModelScore
1GPT-5.263.3
2Gemini 3 Pro (Preview)59.2
3Claude Sonnet 4.554.5
4Qwen 3 VL 235B A22B54.3
5GPT-4.150.7
6Qwen 3 VL 8B50.6
7GPT-4.1 Nano47.8
8Gemini 2.0 Flash43.5

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