PlanBench-V - Caption: leaderboard

Metric: Mean score (out of 2) on map captioning, perception level; answers to expert-curated questions on Chinese territorial spatial planning maps are scored 0 to 2 by a GPT-4o-mini judge (temperature 0) against structured reference answers with annotated critical points; objective items use exact match or semantic similarity; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 17 models tracked.

Top models

#ModelScore
1Qwen 3.6 Plus1.95
2GPT-5.41.9
3GPT-4o1.88
4Claude Opus 4.71.83
5InternVL3-8B1.78
6Gemini 2.5 Pro1.78
7Qwen 3.6 Flash1.64
8Qwen 2.5 VL 7B Instruct1.63
9InternVL3-14B1.58
10Kimi K2.61.52
11Qwen 2 VL 7B1.39
12Qwen 2 VL 2B0.93
13GPT-4o Mini0.89

Interactive version: theaggregate.ai/benchmark?slug=planbench-v-caption · How It Works · Data refreshed daily, snapshot 2026-09-29.