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

Multi-modal Reasoning on MathVision (test)

47.7Accuracy (%)

Nanobanana-Pro

Updated 25d ago

Evaluation Results

Method	Links
Nanobanana-Pro 2026.01		47.7
Qwen-Image 2026.01		47
Qwen-Imgcoder 2026.01		46.9
Gemini-ImgCoder 2026.01		46.9
Nanobanana-Pro 2026.01		46.1
Qwen-Image 2026.01		45.9
Gemini-2.0-Flash 2026.02		41.3
Base 2026.01		39
Qwen2.5-VL-72B 2025.12		38.1
SFT-M 2026.02		32.96
SRPO-7B 2025.12		32.9
ThinkOmni 2026.02		32.9
GRPO 2026.02		32.83
VL-Rethinker-7B 2025.12		32.3
SFT-RS 2026.02		32.25
SFT 2026.02		32.11
SFT-M 2026.02		31.92
ThinkOmni 2026.02		31.5
GPT-4o 2026.02		30.4
ReLaX-VL-7B 2025.12		30.2
GRPO 2026.02		29.96
SFT-RS 2026.02		29.83
SFT 2026.02		29.5
ThinkOmni 2026.02		28.2
ReLaX-VL-3B 2025.12		27.6
Vision-R1-7B 2025.12		27.2
ThinkOmni 2026.02		27
MM-Eureka-7B 2025.12		26.9
Omni-R1 2026.02		25.4
HumanOmniV2 2026.02		25.4
Omni-R1 2026.02		25.4
OpenVLThinker-7B 2025.12		25.3
ThinkOmni 2026.02		25.3
Qwen2.5-VL-7B 2025.12		25.1
Qwen2.5-Omni-7B 2026.02		25
R1-VL-7B 2025.12		24.7
MM-Eureka-8B 2025.12		22.2
Kimi-VL-16B 2025.12		21.4
Qwen2.5-VL-3B 2025.12		21.2
ThinkOmni 2026.02		20.2
Intern2.5-VL-8B 2025.12		19.7
Llava-OV-7B 2025.12		18.5
Qwen2.5-Omni-3B 2026.02		18.2
Intern2-VL-8B 2025.12		17.4
Qwen2-VL-7B 2025.12		16.3