Geospatial Reasoning on GeoMMBench (val)
86.5AccuracyHuman Expert
Evaluation results
| Method | Links | |
|---|---|---|
| Human Expert2026.04 | 86.5 | |
| GeoMMAgentInput Modality=Multimodal2026.04 | 86.5 | |
| Qwen3-VL-30BInput Modality=Multimodal, Model Access=Open-source2026.04 | 70.3 | |
| Gemini-1.5-proInput Modality=Multimodal, Model Access=Closed-source2026.04 | 64.8 | |
| Gemini-2.0-flashInput Modality=Multimodal, Model Access=Closed-source2026.04 | 62.2 | |
| LLaVA-NeXT-34BInput Modality=Multimodal, Model Access=Open-source2026.04 | 59.5 | |
| InternVL 2.5-8BInput Modality=Multimodal, Model Access=Open-source2026.04 | 59.5 | |
| GPT-o1Input Modality=Multimodal, Model Access=Closed-source2026.04 | 59.5 | |
| LLaVA-OneVision-7BInput Modality=Multimodal, Model Access=Open-source2026.04 | 56.8 | |
| Qwen2.5-VL-7BInput Modality=Multimodal, Model Access=Open-source2026.04 | 56.8 | |
| Qwen3-VL-4BInput Modality=Multimodal, Model Access=Open-source2026.04 | 56.8 | |
| Qwen3-VL-8BInput Modality=Multimodal, Model Access=Open-source2026.04 | 56.8 | |
| GPT-4vInput Modality=Multimodal, Model Access=Closed-source2026.04 | 56.8 | |
| GPT-4oInput Modality=Multimodal, Model Access=Closed-source2026.04 | 56.8 | |
| Mini-MonkeyInput Modality=Multimodal, Model Access=Open-source2026.04 | 54.1 | |
| Qwen2.5-VL-3BInput Modality=Multimodal, Model Access=Open-source2026.04 | 54.1 | |
| Phi-3.5vInput Modality=Multimodal, Model Access=Open-source2026.04 | 51.4 | |
| Qwen2-VL-7BInput Modality=Multimodal, Model Access=Open-source2026.04 | 51.4 | |
| NVILA-15BInput Modality=Multimodal, Model Access=Open-source2026.04 | 48.6 | |
| Phi-4-MultimodalInput Modality=Multimodal, Model Access=Open-source2026.04 | 48.6 | |
| LLaVA-NeXT-7BInput Modality=Multimodal, Model Access=Open-source2026.04 | 45.9 | |
| Phi-3v-128kInput Modality=Multimodal, Model Access=Open-source2026.04 | 45.9 | |
| InternLM-XComposer2-4KHDInput Modality=Multimodal, Model Access=Open-source2026.04 | 44.5 | |
| NVILA-8BInput Modality=Multimodal, Model Access=Open-source2026.04 | 43.2 | |
| VHMInput Modality=Multimodal, Model Access=Open-source2026.04 | 37.8 | |
| CogVLM2Input Modality=Multimodal, Model Access=Open-source2026.04 | 37.8 | |
| InternLM-XComposer-2.5-7BInput Modality=Multimodal, Model Access=Open-source2026.04 | 37.8 | |
| LLaMA3.2-11BInput Modality=Multimodal, Model Access=Open-source2026.04 | 35.1 | |
| GPT-5Input Modality=Multimodal, Model Access=Closed-source2026.04 | 35.1 | |
| LLaVA-OneVision-0.5BInput Modality=Multimodal, Model Access=Open-source2026.04 | 32.4 | |
| LLaVA1.5-13BInput Modality=Multimodal, Model Access=Open-source2026.04 | 29.7 | |
| Qwen2-VL-2BInput Modality=Multimodal, Model Access=Open-source2026.04 | 29.7 | |
| LHRS-BotInput Modality=Multimodal, Model Access=Open-source2026.04 | 27 | |
| LLaVA1.5-7BInput Modality=Multimodal, Model Access=Open-source2026.04 | 27 | |
| GPT-4 TextInput Modality=Text-only, Model Access=Closed-source2026.04 | 24.3 | |
| TeoChatInput Modality=Multimodal, Model Access=Open-source2026.04 | 24.3 | |
| GeoChatInput Modality=Multimodal, Model Access=Open-source2026.04 | 21.6 | |
| Random2026.04 | 20.8 | |
| SkySenseGPTInput Modality=Multimodal, Model Access=Open-source2026.04 | 18.9 |