GeoMMBench - Principles: leaderboard

The principles and theory questions of GeoMMBench (2026), an expert-written multimodal geoscience and remote sensing question benchmark answered zero-shot (test set). Micro-averaged accuracy in percent; higher is better.

Metric: Micro-averaged accuracy (%) on the principles and theoretical understanding questions of the GeoMMBench test set (expert-written multimodal geoscience and remote sensing multiple-choice and numeric questions, zero-shot, LMMs-Eval, rule-based answer extraction, invalid outputs counted wrong); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 35 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 1.5 Pro71.5#1071
2Gemini 2.0 Flash68.9#731
3GPT-4o65.8#754
4O164.2#407
5Qwen 3 VL 30B63.2#566
6Qwen 3 VL 8B61.4#927
7Qwen 2.5 VL 7B59.8#1878
8GPT-4V59.8#1483
9Qwen 3 VL 4B58.2#1163
10Qwen 2.5 VL 3B54.3#2664
11Qwen 2 VL 7B53.8#2300
12LLaVA-OneVision-7B53.5#3002
13Phi-4 Multimodal Instruct53.3#4155
14LLaVA-NeXT-34B50.7#3483
15InternVL2.5-8B50.4#2393

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=geommbench-principles · How It Works · Data refreshed daily, snapshot 2026-11-22.