List your company
ResearchFormerly WizWand

Visual Programming and Reasoning on HumanEval_V

28.5AccuracyKimi-K2.5
0.1087.47914.8522.221May 21, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Kimi-K2.5Model Category=Open-Source & Baselines2026.0528.5
MAESTRO*Evaluation Protocol=augments the registry with 2 additional experts and 4 new Level-1 skills2026.0528.1
GLM-4.6VModel Category=Open-Source & Baselines2026.0527.7
GPT-5Model Category=Closed-Source Models2026.0525.3
Gemini-2.5-FlashModel Category=Closed-Source Models2026.0523.7
GPT-4oModel Category=Closed-Source Models2026.0523.3
Gemini-2.5-ProModel Category=Closed-Source Models2026.0521.7
MAESTROEvaluation Protocol=default pool2026.0519.4
Best ModelModel Category=Open-Source & Baselines2026.0513.3
Qwen3-VL-32BModel Category=Open-Source & Baselines2026.055.5
Direct AnsweringModel Category=Open-Source & Baselines2026.054
DeepEyesModel Category=Think with Images Methods2026.054
VTOOL-R1Model Category=Think with Images Methods2026.054
VisionReasonerModel Category=Think with Images Methods2026.054
Visual-ARFTModel Category=Think with Images Methods2026.052.8
Untrained ModelModel Category=Open-Source & Baselines2026.052.7
VTS-VModel Category=Think with Images Methods2026.052.4
DeepEyes-v2Model Category=Think with Images Methods2026.051.6
PixelReasonerModel Category=Think with Images Methods2026.051.6
Chain-of-FocusModel Category=Think with Images Methods2026.051.6
ThymeModel Category=Think with Images Methods2026.051.2
MathCoder-VLModel Category=Think with Images Methods2026.051.2