AtelierEval - Image Constraint Satisfaction (Novice Prompting): leaderboard
AtelierEval (Luo et al., 2026) tests models as text-to-image prompters over 360 expert-written tasks. Each prompt is rendered four times by each of four image generators (Gemini 3 Pro Image, GPT-Image-1, Flux.1 Pro, SDXL), the best image per generator is kept, and a GPT-5.4 judge scores the share of the constraint checklist of the task visible in it, averaged over generators and task categories, in percent (higher is better). The model receives a direct natural-language request to write the prompt.
Metric: Image checklist satisfaction (%; share of each task's expert constraint checklist realized in the generated image, best of four images per backend, mean over four text-to-image backends (Gemini 3 Pro Image, GPT-Image-1, Flux.1 Pro, SDXL) and the three task categories, GPT-5.4 AtelierJudge; direct natural-language prompting). Source: arxiv.org. Saturation forecast: Around November 2028. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 65.7 | #172 |
| 2 | Gemini 3 Pro (Preview) | 64.7 | #111 |
| 3 | Claude Sonnet 4.5 | 64.1 | #244 |
| 4 | GPT-4.1 | 63 | #492 |
| 5 | Qwen 3 VL 235B A22B | 61.3 | #477 |
| 6 | Gemini 2.0 Flash | 60.9 | #702 |
| 7 | GPT-4.1 Nano | 60.8 | #1992 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=ateliereval-image-constraint-satisfaction-novice-prompting · How It Works · Data refreshed daily, snapshot 2026-11-12.