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

Vision-Language Reasoning on DisasterBench (val)

87.7Overall Accuracy

Gemini-2.5-pro

Updated 1mo ago

Evaluation Results

Method	Links
Gemini-2.5-pro 2026.06		87.7	1,323
Claude-Sonnet-4-5 2026.06		86.26	457
GPT-5 2026.06		83.69	644
Gemini-2.5-flash 2026.06		82.96	1,321
Gemini-2.0-flash 2026.06		81.68	435
DisasterVL 2026.06		81.57	168
Grok-4 2026.06		77.87	733
Qwen2.5-VL-7B 2026.06		74.94	425
GPT-4o 2026.06		73.19	335
InternVL-3-5-4B 2026.06		72.41	459
InternVL-3-5-8B 2026.06		72.31	322
InternVL2-8B 2026.06		71.38	255
Deepseek-VL2-S 2026.06		70.97	359
GPT-4o mini 2026.06		67.27	423
Phi-3-Vision(4.15B) 2026.06		66.8	272
InternVL-3-5-2B 2026.06		66.13	355
Qwen2.5-VL-3B 2026.06		65.31	428
InternVL-3-5-1B 2026.06		55.79	380
Janus-Pro-1B 2026.06		55.12	354
InternVL2-2B 2026.06		54.09	425
Deepseek-VL2-T 2026.06		52.91	190
Qwen2-VL-2B 2026.06		49.92	352