Share your thoughts, 1 month free Claude Pro on usSee more

Geospatial Reasoning on GeoMMBench (val)

86.5Accuracy

Human Expert

Updated 1mo ago

Evaluation Results

Method	Links
Human Expert 2026.04		86.5
GeoMMAgent 2026.04		86.5
Qwen3-VL-30B 2026.04		70.3
Gemini-1.5-pro 2026.04		64.8
Gemini-2.0-flash 2026.04		62.2
LLaVA-NeXT-34B 2026.04		59.5
InternVL 2.5-8B 2026.04		59.5
GPT-o1 2026.04		59.5
LLaVA-OneVision-7B 2026.04		56.8
Qwen2.5-VL-7B 2026.04		56.8
Qwen3-VL-4B 2026.04		56.8
Qwen3-VL-8B 2026.04		56.8
GPT-4v 2026.04		56.8
GPT-4o 2026.04		56.8
Mini-Monkey 2026.04		54.1
Qwen2.5-VL-3B 2026.04		54.1
Phi-3.5v 2026.04		51.4
Qwen2-VL-7B 2026.04		51.4
NVILA-15B 2026.04		48.6
Phi-4-Multimodal 2026.04		48.6
LLaVA-NeXT-7B 2026.04		45.9
Phi-3v-128k 2026.04		45.9
InternLM-XComposer2-4KHD 2026.04		44.5
NVILA-8B 2026.04		43.2
VHM 2026.04		37.8
CogVLM2 2026.04		37.8
InternLM-XComposer-2.5-7B 2026.04		37.8
LLaMA3.2-11B 2026.04		35.1
GPT-5 2026.04		35.1
LLaVA-OneVision-0.5B 2026.04		32.4
LLaVA1.5-13B 2026.04		29.7
Qwen2-VL-2B 2026.04		29.7
LHRS-Bot 2026.04		27
LLaVA1.5-7B 2026.04		27
GPT-4 Text 2026.04		24.3
TeoChat 2026.04		24.3
GeoChat 2026.04		21.6
Random 2026.04		20.8
SkySenseGPT 2026.04		18.9