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

Visual Programming and Reasoning on HumanEval_V

28.5Accuracy

Kimi-K2.5

Updated 12d ago

Evaluation Results

Method	Links
Kimi-K2.5 2026.05		28.5
MAESTRO* 2026.05		28.1
GLM-4.6V 2026.05		27.7
GPT-5 2026.05		25.3
Gemini-2.5-Flash 2026.05		23.7
GPT-4o 2026.05		23.3
Gemini-2.5-Pro 2026.05		21.7
MAESTRO 2026.05		19.4
Best Model 2026.05		13.3
Qwen3-VL-32B 2026.05		5.5
Direct Answering 2026.05		4
DeepEyes 2026.05		4
VTOOL-R1 2026.05		4
VisionReasoner 2026.05		4
Visual-ARFT 2026.05		2.8
Untrained Model 2026.05		2.7
VTS-V 2026.05		2.4
DeepEyes-v2 2026.05		1.6
PixelReasoner 2026.05		1.6
Chain-of-Focus 2026.05		1.6
Thyme 2026.05		1.2
MathCoder-VL 2026.05		1.2