List your company

Earth-science reasoning on MSEarthMCQ

61.8AccuracyGPT-5
35.842.5549.356.05May 21, 2026

Evaluation results

MethodLinks
GPT-5Model Type=Closed-Source2026.0561.8
Gemini-2.5-ProModel Type=Closed-Source2026.0555.8
Kimi-K2.5Model Type=Open-Source & Baselines2026.0555
GLM-4.6VModel Type=Open-Source & Baselines2026.0554.2
Gemini-2.5-FlashModel Type=Closed-Source2026.0554
MAESTROModel Type=Ours2026.0552.4
Qwen3-VL-32BModel Type=Open-Source & Baselines2026.0551
Visual-ARFTModel Type=Think with Images Methods2026.0546.6
DeepEyes-v2Model Type=Think with Images Methods2026.0546.4
PixelReasonerModel Type=Think with Images Methods2026.0546.2
VisionReasonerModel Type=Think with Images Methods2026.0545.8
VTOOL-R1Model Type=Think with Images Methods2026.0545.4
VTS-VModel Type=Think with Images Methods2026.0545.4
DeepEyesModel Type=Think with Images Methods2026.0545
Chain-of-FocusModel Type=Think with Images Methods2026.0544.2
MathCoder-VLModel Type=Think with Images Methods2026.0543.8
ThymeModel Type=Think with Images Methods2026.0542.2
GPT-4oModel Type=Closed-Source2026.0542
Direct AnsweringModel Type=Open-Source & Baselines2026.0539.8
Untrained ModelModel Type=Open-Source & Baselines2026.0536.8