List your company

Spatial Reasoning on VSI-Bench Vanilla regime

50.5Avg ScoreGeoThinker Qwen2.5VL-7B
27.72433.63739.5545.463Feb 5, 2026

Evaluation results

MethodLinks
GeoThinker Qwen2.5VL-7BBackbone=Qwen2.5VL-7B, Active Perception=true2026.0250.569.538.557.962.245.246.231.452.6
VG-LLM-8BModel Category=Open-Source Spatial Intelligence Model, Parameters=8B, Active Perception=false2026.0249.768.138.75961.145.544.926.853.4
GeoThinker Qwen2.5VL-3BBackbone=Qwen2.5VL-3B, Active Perception=true2026.0248.968.536.157.362.543.747.934.540.9
Spatial-MLLM-4BModel Category=Open-Source Spatial Intelligence Model, Parameters=4B, Active Perception=false2026.0248.465.334.863.145.141.346.233.546.3
VG-LLM-4BModel Category=Open-Source Spatial Intelligence Model, Parameters=4B, Active Perception=false2026.0246.767.637.655.252.54844.731.935.5
Gemini-1.5 ProModel Category=Proprietary Model (API)2026.0245.456.230.964.143.651.346.33634.6
SpatialLadder-3BModel Category=Open-Source Spatial Intelligence Model, Parameters=3B2026.0244.8————————
SPAR-8BModel Category=Open-Source Spatial Intelligence Model, Parameters=8B2026.0244.1————————
Gemini-1.5 FlashModel Category=Proprietary Model (API)2026.0242.149.830.853.554.437.74131.537.8
LLaVA-NeXT-Video-72BModel Category=Open-Sourced Model, Parameters=72B2026.0240.948.922.857.435.342.436.73548.6
LLaVA-OneVision-72BModel Category=Open-Sourced Model, Parameters=72B2026.0240.243.523.957.637.542.539.932.544.6
InternVL2-40BModel Category=Open-Sourced Model, Parameters=40B2026.023634.926.946.531.842.132.23439.6
LLaVA-NeXT-Video-7BModel Category=Open-Sourced Model, Parameters=7B2026.0235.648.51447.824.243.542.43430.6
InternVL2-8BModel Category=Open-Sourced Model, Parameters=8B2026.0234.623.128.748.239.836.730.729.939.6
Chance Level (Frequency)Type=Baseline2026.023462.13229.933.125.147.928.425.2
GPT-4OModel Category=Proprietary Model (API)2026.023446.25.343.838.23741.331.528.5
LLaVA-OneVision-7BModel Category=Open-Sourced Model, Parameters=7B2026.0232.447.720.247.412.342.535.229.424.4
Qwen2.5VL-7BModel Category=Open-Sourced Model, Parameters=7B2026.0229.325.210.536.429.638.43829.826.8
Qwen2.5VL-3BModel Category=Open-Sourced Model, Parameters=3B2026.0228.632.719.517.325.137.344.930.421.8
Chance Level (Random)Type=Baseline2026.02—————2536.128.325