PlanBench-V - Classification: leaderboard

Metric: Mean score (out of 2) on classifying planning maps by type within the statutory planning system, reasoning 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: Around December 2026. 17 models tracked.

Top models

#ModelScore
1Qwen 3.6 Plus1.67
2Gemini 2.5 Pro1.66
3GPT-5.41.56
4GPT-4o1.41
5Kimi K2.61.34
6Claude Opus 4.71.32
7Qwen 3.6 Flash1.3
8InternVL3-14B1.18
9Qwen 2.5 VL 7B Instruct1.09
10Qwen 2 VL 7B1.03
11InternVL3-8B1.03
12GPT-4o Mini1.02
13Qwen 2 VL 2B0.95

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