List your company

Spatial Reasoning on VSI-Bench

95.8R.Dr.Human
45.8858.8471.884.76May 26, 2025Aug 1, 2025Oct 8, 2025Dec 15, 2025Feb 21, 2026Apr 30, 2026Jul 7, 2026

Evaluation results

MethodLinks
HumanCategory=Baseline2026.0295.84760.494.794.310079.2—95.845.9———————————
HumanDescription=Reported results on a subset of 400 samples2026.0395.8——94.7—10079.2—95.845.994.360.4—47———————
HumanEvaluation Protocol=EASI protocol [6]2026.0495.847—94.7——79.2—95.845.994.360.4———100—————
Human2026.0695.8——94.7—10079.2—95.845.994.360.4—47———————
HumanCategory=Reference2026.0695.847—94.7—10079.2—95.845.994.360.4—————————
SSR-3DCategory=Open-Sourced Spatial Models2026.0293.467.17671.370.58573.9—48.579.5———————————
Cambrian-PLM=Qwen2.5-7B2026.0589.5——74.8—8573.7—52.676.974.976—60.1———————
GS-Reasoner (pred dep.)Category=Open-Sourced Spatial Models2026.0288.961.97065.469.152.364.7—44.365.7———————————
SpaceMind-8Bw/o E3D=false2026.0688.4——67.2—70.669.6—44.374.273.377.3—61.4———————
SpaceMind2026.0688.461.4—67.2—70.669.6—44.374.273.377.3—————————
SSR-2DCategory=Open-Sourced Spatial Models2026.0287.963.476.569.271.881.971.9—48.576.5———————————
VLM²-7Bparameters=7B2026.0587.859.6—69—68.368.8—52.669.972.570.8—————————
Spatial Code-4BCategory=Open-Source Spatial Intelligence Models, Model Size=4B, Additional input=2D bounding box annotations2026.0687.160.7—62—5960—32.533.19250.8—————————
Q-GeoMembackbone=LLaVA-NeXT-Video-7B, fine-tuning=LoRA2026.0585.554.5—71.4—73.470.1—46.976.376.376.7—————————
OneCanvas2026.0684.859.5—71—65.770.1—59.876.568.475.4—————————
SpatialStack-5BRank=1, Type=Open-source Models, Backbone=Qwen3.52026.0384.155.6—67.3—83.567.5—41.268.27169.1—————————
SpatialStack-5Bw/o E3D=false2026.0684.1——67.3—83.567.5—41.268.27169.1—55.6———————
Cambrian-S-7B†LM=Qwen2.5-7B, fine-tuned only on VSI-590K=true2026.0582——71.5—84.369.2—38.774.773.675.2—53.7———————
SenseNova-SI2026.0680.853.5—69.6—76.468.8—48.572.87276.8—————————
SenseNova-SI InternVL3-8BCategory=Open-Sourced Spatial Models2026.0280.77272.776.376.748.468.7—69.553.5———————————
GeoVR-2Bw/o E3D=true2026.0680.7——71.3—86.769.1—45.972.367.773.9—54.5———————
VLM-3R-7BCategory=Open-Sourced Spatial Models2026.0280.549.469.265.470.240.160.9—45.467.1———————————
VLM-3R (7B)Parameters=7B, Finetuned=true2025.0580.5——65.4—40.160.9—45.467.170.269.2—49.4———————
VLM-3R-7BEvaluation Protocol=EASI protocol [6]2026.0480.549.4—65.4——60.9—45.467.170.269.2———40.1—————
VLM-3R 7BLM=Qwen2-7B2026.0580.5——65.4—40.160.9—45.467.170.269.2—49.4———————
VLM-3R-7Bparameters=7B2026.0580.549.4—65.4—40.160.9—45.467.170.269.2—————————
VLM-3RModel Category=3D Spatial Reasoning VLMs, Training Data=Finetuning on 3D Related VQAs2026.0580.5———————45.4—70.2——————————
VLM-3R-7Bw/o E3D=false2026.0680.5——65.4—40.160.9—45.467.170.269.2—49.4———————
VLM-3R-7BCategory=Open-Source Spatial Intelligence Models, Model Size=7B2026.0680.549.4—65.4—40.160.9—45.467.170.269.2—————————
VLM-3R2026.0680.549.4—65.4—40.160.9—45.467.170.269.2—————————
Cambrian-S-7BCategory=Open-Sourced Spatial Models2026.0276.250.574.971.173.280.167.5—41.872.2———————————
Cambrian-S 7BLM=Qwen2.5-7B2026.0576.2——71.1—80.167.5—41.872.273.274.9—50.5———————
Cambrian-S-7Bw/o E3D=true2026.0676.2——71.1—80.167.5—41.872.273.274.9—50.5———————
Cambrian-S2026.0676.250.5—71.1—80.167.5—41.872.273.274.9—————————
Spatial-TTT-2BCategory=Ours, Size=2B2026.0373——61.8—7764.4—47.465.970.871.7—47.8———————
SpaR3D-MoESparse frames=32, Model Category=Spatial-Aware MLLMs2026.0770.148—58.7—73.563.5—4469.671.372.8—————————
SpatialImaginerBackbone=Bagel-7B-MoT, Evaluation Protocol=EASI protocol [6]2026.0469.946.8—67.4——63.7—46.464.670.672.7———71.2—————
Cambrian-S-7BEvaluation Protocol=EASI protocol [6]2026.0469.645.8—66.8——62.9—39.267.668.272.5———73.8—————
Cambrian-S-7BCategory=Open-Source Spatial Intelligence Models, Model Size=7B, Evaluation frames=322026.0669.645.8—66.8—73.862.9—39.267.668.272.5—————————
LLaVA-Next-Video-7B (Finetuned)Parameters=7B, Finetuned=true2025.0568.9——64.9—38.557.7—40.763.770.670.8—43.6———————
SpatialStack-4BRank=2, Type=Open-source Models, Backbone=Qwen2.52026.0368.445.4—57.9—79.660.9—40.263.269.263—————————
SpatialStack-4Bw/o E3D=false2026.0668.4——57.9—79.660.9—40.263.269.263—45.4———————
Skill-3DModel=GPT-5.42026.0667——62.7—78.1——71.477.666.274.9—61.5———————
LLaVA-One-Vision-7B (Finetuned)Parameters=7B, Finetuned=true2025.0565.9——59.7—39.455.8—41.759.568.366.5—45.3———————
Skill-3DModel=Gemini-2.5-Pro2026.0664.2——67.6—76.4——6972.862.473.1—58———————
Ouro-Spatial-4BCategory=Open-Source Spatial Intelligence Models, Model Size=4B2026.0663.747.5—63.8—71.462.7—3373.571.577.1—————————
Skill-3DModel=Gemini-3-Flash2026.0663——60.4—73.4——67.571.860.971.2—56.1———————
Ouro-Spatial-8BCategory=Open-Source Spatial Intelligence Models, Model Size=8B2026.0662.648.7—64.2—70.763.3—35.675.771.777.3—————————
GR3DCombined Model=GPT-5-2025-08-072026.0362.2——57.6—68.158.7—61.260.753.463.5—42.7———————
Cambrian-S-3BCategory=Open-source Spatial Intelligence Models, Size=3B2026.0361.9——64.8—78.857.3—27.346.370.768—40.6———————
Cambrian-S-3BRank=3, Type=Open-source Models2026.0361.940.6—64.8—78.857.3—27.346.370.768—————————
Cambrian-S-3Bw/o E3D=true2026.0661.9——64.8—78.857.3—27.346.370.768—40.6———————
Cambrian-S-3BCategory=Open-Source Spatial Intelligence Models, Model Size=3B, Evaluation frames=322026.0661.940.6—64.8—78.857.3—27.346.370.768—————————
Think3DModel=GPT-5.42026.0661.8——58.3—73.8——66.569.961.170.8—55———————
Think3DModel=Gemini-3-Flash2026.0660.8——56.5—69.3——64.266.356.868—52.3———————
Skill-3D-8BModel=Qwen3-VL-8B2026.0660.148.6—52—66.8——58.467.956.559.8—————————
Think3DModel=Gemini-2.5-Pro2026.0659.5——64.8—72.2——65.866.458.269.5—53.1———————
Skill-3DModel=GPT-4o2026.0659.2——53.4—35.4——62.769.556.858.1—42.6———————
Gemini-3-proCategory=Proprietary Models2026.0357.5——56.6—6056—61.941.84971.5—42.8———————
Qwen3.5-4BRank=4, Type=Open-source Models2026.0357.536.5—60.3—62.353.6—3453.856.567.5—————————
Qwen3.5-4B2026.0657.5——60.3—62.353.6—3453.856.567.5—36.5———————
Gemini-3-ProCategory=Proprietary Models2026.0657.542.8—56.6—6856—61.941.84971.5—————————
ReMoT -CoT-4BMethod Variation=Ours2026.0257.442.5—52.1—59.758.8—45.656.378.276.81————————
VST-7B-SFTCategory=Open-source Spatial Intelligence Models, Size=7B2026.0355.8——59.7—65.260.6—44.968.37274.3—44.4———————
VST-7Bparameters=7B2026.0555.844.4—59.7—65.260.6—44.968.37274.3—————————
VST-7B-SFTw/o E3D=true2026.0655.8——59.7—65.260.6—44.968.37274.3—44.4———————
VST-7B-SFTCategory=Open-Source Spatial Intelligence Models, Model Size=7B2026.0655.844.4—59.7—65.260.6—44.968.37274.3—————————
Gemini-3-ProEvaluation Protocol=EASI protocol [6]2026.0455.737.8—59.9——52.5—45.944.13872.7———66—————
VST-7BCategory=Open-Sourced Spatial Models2026.0255.643.875.56071.669.261.2—44.369.2———————————
VST 7BLM=Qwen2.5-7B2026.0555.6——60—69.261.2—44.369.271.675.5—43.8———————
VST2026.0655.643.8—60—69.261.2—44.369.271.675.5—————————
w/ ToolsModel=GPT-5.42026.0653.3——51.7—74.6——52.558.25869.3—47.8———————
Think3D-8BModel=Qwen3-VL-8B2026.0652.938.5—41.6—60.8——45.458.148.351.2—————————
Think3DModel=GPT-4o2026.0652.6——47.9—29.2——55.961.450.450.5—32.7———————
Qwen3-VL-8BCategory=Open-Source General Models, Model Size=8B2026.0651.547.1—58.3—64.256.5—31.457.26675.9—————————
w/ ToolsModel=Gemini-2.5-Pro2026.0651.2——62.6—72.7——49.546.348.466—41.5———————
Qwen3-VL-8B-InstructEvaluation Protocol=EASI protocol [6]2026.045147—58——57.9—3561.967.676.3———66.3—————
Qwen3-VL-8B-InstructCategory=Open-Sourced General Models2026.0250.94776.35853.366.357.9—3561.9———————————
Qwen3-VL-8B-InstructCategory=Open-source General Models, Size=8B2026.0350.9——58—66.357.9—3561.967.576.3—47———————
Qwen3-VL-8B-Instruct2026.0650.9——58—66.357.9—3561.967.576.3—47———————
w/o ToolsModel=GPT-5.42026.0650.5——48.4—73.2——49.155.755.867.3—43.6———————
Skill-3D-4BModel=Qwen3-VL-4B2026.0650.436.8—43.5—56.7——48.857.448.650.2—————————
Robust-TO (Qwen3-VL-7B-Instruct)Frames=20.72026.0648.8——55.5——56.4—39.8———————————77.5
GPT-5Evaluation Protocol=EASI protocol [6]2026.0448.734.5—63.7——55—50.347.553.373.3———68.9—————
VST-7B-SFTEvaluation Protocol=EASI protocol [6]2026.0448.737.3—55.2——55.5—41.862.268.874.5———55.5—————
GPT-5-2025-08-07Category=Proprietary Models2026.0248.634.473.363.753.568.955—50.247.5———————————
GPT-5Category=Proprietary Models2026.0348.6——63.7—68.955—50.247.553.373.3—34.4———————
GPT-52026.0648.6——63.7—68.955—50.247.553.373.3—34.4———————
GPT-5Category=Proprietary Models2026.0648.634.4—47.8—68.955—50.248.353.373.3—————————
w/ ToolsModel=Qwen3-VL-8B2026.0648.427.8—44.7—52.6——43.24644.748.1—————————
Spatial-MLLM-4BEvaluation Protocol=EASI protocol [6]2026.0448.238.1—40.4——46.3—3335.566.763.6———44.3—————
Gemini-1.5-ProModel Category=Proprietary Models (API)2026.0348.128.8—46—6848.8—4249.449.658.6—————————
Random Choice2026.0347.9——25.1—25.234—28.433.162.129.9—32———————
Chance Level (Frequency)2026.0347.932—25.1—25.234—28.433.162.129.9—————————
Chance Level (Frequency)Type=Baseline2025.0547.9——25.1—25.234—28.433.162.129.9—32———————
Random Choice(Frequency)Evaluation Protocol=EASI protocol [6]2026.0447.932—25.1——34—28.433.162.129.9———25.2—————
Chance Level (Frequency)2026.0547.9——25.1—25.234—28.433.162.129.9—32———————
RandomCategory=Reference2026.0647.932—25.1—25.234—28.433.162.129.9—————————
Gemini-2.5 ProRank=1, Type=Proprietary Models (API)2026.0347.834.9—61.1—71.351.5—45.942.843.864.3—————————
Gemini-2.5 ProType=Proprietary Model (API)2025.0547.8——61.1—71.351.5—45.942.843.864.3—34.9———————