List your company

Spatial Reasoning on Escher-Bench 1.0 (test)

56.17Object Permanence & Occlusion TrackingGemini-2.5-pro
4.263617.739331.21544.6907Jan 4, 2026

Evaluation results

MethodLinks
Gemini-2.5-proAccess=Closed-Source, Paradigm=API2026.0156.1757.2969.5354.273.1154.3957.2658.3753.6464.5359.651.5760.74
Gemini-2.5-flashAccess=Closed-Source, Paradigm=API2026.0153.1351.7165.0149.2265.8848.9152.1952.9349.6656.0248.5545.5158.42
GPT5Access=Closed-Source, Paradigm=API2026.0152.2453.2864.6650.4167.9950.5853.3654.2849.8860.0155.4347.9656.49
Escher-4B-InstructAccess=Open-Source, Paradigm=SFT, Size=4B2026.0149.7248.5961.6347.3352.6147.7149.4949.7348.7151.8241.874752.41
Qwen3-VL-32B-ThinkingAccess=Open-Source, Paradigm=Thinking, Size=32B2026.0149.2649.858.7649.3460.1946.0449.5849.749.1857.6848.5543.353.87
Qwen3-VL-30B-A3B-InstructAccess=Open-Source, Paradigm=Instruct, Architecture=MoE, Size=30B2026.0148.5344.4654.3843.0856.8743.745.7346.3543.7547.7937.8639.0453.05
Escher-4B-ThinkingAccess=Open-Source, Paradigm=SFT / Thinking, Size=4B2026.0148.0747.5561.3347.3253.0846.3848.5448.8647.4949.7243.6545.2152.26
Escher-7B-InstructAccess=Open-Source, Paradigm=SFT, Size=7B2026.0147.6148.8862.0848.8953.5545.3148.8949.1747.9751.0541.246.8851.38
Escher-8B-InstructAccess=Open-Source, Paradigm=SFT, Size=8B2026.0146.9649.863.648.3156.8747.549.8550.8846.5552.648.1146.0853.17
Qwen3-VL-8B-InstructAccess=Open-Source, Paradigm=Instruct, Size=8B2026.0146.6943.9855.8943.0858.2942.2145.0645.7442.8647.7938.3137.8252.6
ISI-8B-InstructAccess=Open-Source, Paradigm=SFT, Size=8B2026.0146.5943.0156.543.4650.2443.6445.0646.0641.8650.6126.538.9252.32
Escher-8B-ThinkingAccess=Open-Source, Paradigm=SFT / Thinking, Size=8B2026.0146.0448.6762.8448.6460.1945.2248.7949.1247.7150.2844.9946.0851.66
Qwen3-VL-4B-InstructAccess=Open-Source, Paradigm=Instruct, Size=4B2026.0144.5743.7448.494352.6143.0844.1844.941.8640.8835.6338.7451.78
qwen3-vl-plusAccess=Closed-Source, Paradigm=API2026.0144.3843.2155.5244.6157.8242.3844.7545.8341.2844.8644.7738.6350.94
MiniCPM-V-4Access=Open-Source, Paradigm=Instruct2026.0143.1938.6746.5337.9432.2338.5439.6239.8139.0635.2522.2736.6546.22
Qwen3-VL-4B-ThinkingAccess=Open-Source, Paradigm=Thinking, Size=4B2026.0143.0943.3249.8543.6254.544.6344.5546.0139.8552.1535.1939.0449.35
LLaVA-OneVision-1.5-4B-InstructAccess=Open-Source, Paradigm=Instruct, Size=4B2026.0142.8243.3548.6441.5250.7143.6443.7344.2741.9644.6436.338.5649.74
LLaVA-OneVision-1.5-8B-InstructAccess=Open-Source, Paradigm=Instruct, Size=8B2026.0141.7144.8748.9442.5949.2944.6744.545.2242.1742.8736.7539.3451.26
ERNIE-4.5-VL-28B-A3B-ThinkingAccess=Open-Source, Paradigm=Thinking, Architecture=MoE, Size=28B2026.0141.3440.1751.6640.2857.3541.2242.5843.1141.0849.7235.1937.8549.01
Qwen3-VL-8B-ThinkingAccess=Open-Source, Paradigm=Thinking, Size=8B2026.0141.2544.2651.2141.0258.2944.4544.3645.740.0651.8241.4338.6248.56
Qwen2.5-VL-7B-InstructAccess=Open-Source, Paradigm=Instruct, Size=7B2026.0139.3235.7345.1738.0238.3933.7536.8436.6937.3226.0812.9232.3947.59
Qwen2.5-VL-3B-InstructAccess=Open-Source, Paradigm=Instruct, Size=3B2026.0139.1338.9847.1341.8944.5540.1540.6141.1238.9636.4111.3635.4650.99
Qwen3-VL-2B-InstructAccess=Open-Source, Paradigm=Instruct, Size=2B2026.0137.7537.5344.1138.3541.2338.6338.6539.6235.5331.717.834.2749.23
Spatial-MLLMAccess=Open-Source2026.0110.615.226.1322.811.3712.8115.8215.616.5516.021.1133.280
ViCA2-7BAccess=Open-Source, Size=7B2026.019.6715.2427.7923.9714.6913.9216.4515.9618.0317.57234.210.03
LLaVA-NeXT-Video-7BAccess=Open-Source, Paradigm=Instruct, Size=7B2026.016.2610.288.9116.052.8413.6611.310.6313.441.16.4625.80