List your company

Video Understanding on Perception (test)

82.7AccuracyPLM-8B
31.53244.81658.171.384Feb 5, 2024Jun 2, 2024Sep 28, 2024Jan 25, 2025May 23, 2025Sep 18, 2025Jan 15, 2026

Evaluation results

MethodLinks
PLM-8BAccess Type=Open models2026.0182.7—
Molmo2-8BAccess Type=Molmo2 family2026.0182.1—
Molmo2-4BAccess Type=Molmo2 family2026.0181.3—
Eagle2.5-8BAccess Type=Open weights only2026.0181—
Molmo2-O-7BAccess Type=Molmo2 family2026.0179.6—
GPT-5Access Type=API call only2026.0179.4—
PLM-3BAccess Type=Open models2026.0179.3—
Gemini 2.5 ProAccess Type=API call only2026.0178.4—
Gemini 3 ProAccess Type=API call only2026.0177.6—
VideoChat-Flash-7BAccess Type=Open models2026.0176.5—
Gemini 2.5 FlashAccess Type=API call only2026.0174.7—
GLM-4.1V-9BAccess Type=Open weights only2026.0174.2—
InternVL3.5-8BAccess Type=Open weights only2026.0172.7—
Qwen3-VL-8BAccess Type=Open weights only2026.0172.7—
GPT-5 miniAccess Type=API call only2026.0172—
MiniCPM-V-4.5-8BAccess Type=Open weights only2026.0170.9—
Qwen3-VL-4BAccess Type=Open weights only2026.0170.7—
LLaVA-Video-7BAccess Type=Open models2026.0168.8—
InternVL3.5-4BAccess Type=Open weights only2026.0168.1—
Qwen2-VL-72BParameters=72B2024.0968—
Previous SoTA2024.0966.9—
LLaVA-OV-72BBackbone=72B2024.1266.9—
Claude Sonnet 4.5Access Type=API call only2026.0164.3—
Keye-VL-1.5-8BAccess Type=Open weights only2026.0164.2—
Qwen2-VL-7BParameters=7B2024.0962.3—
LLaVA-OV-72Bvariant=Single-Image (SI), Backbone=72B2024.1262.3—
Qwen2-VL-7B-Ins.Backbone=7B-Instruct2024.1262.3—
Llama 3-V 70BZero-shot=true, Parameters=70B2024.0760.8—
MAmmoTH-VL-8BBackbone=8B2024.1259.3—
InternVL-2-8BBackbone=8B2024.1257.4—
LLaVA-OV-7BBackbone=7B2024.1257.1—
LLaVA-OV-7Bvariant=Single-Image (SI), Backbone=7B2024.1254.9—
Gemini 1.0 UltraZero-shot=true2024.0754.7—
Qwen2-VL-2BParameters=2B2024.0953.9—
Llama 3-V 8BZero-shot=true, Parameters=8B2024.0753.8—
Gemini 1.0 ProZero-shot=true2024.0751.1—
Video-LaVITZero-shot=true2024.0247.9—
VideoChat2Zero-shot=true2024.0247.3—
BLIP-2Zero-shot=true2024.0239.2—
FlamingoZero-shot=true2024.0233.5—
BaselineModel=Baseline, Pruning Strength=0%2026.03—78.7
CASA⊕ Q2.5VLMax Res.=448^22025.12—58.1
InternVL3.5-8BModel=InternVL3.5-8B, Pruning Strength=None2026.03—72.7
PLM-8BModel=PLM-8B, Pruning Strength=None2026.03—82.7
Qwen2.5VL 3BMax Res.=Native2025.12—67
Qwen3-VL-4BModel=Qwen3-VL-4B, Pruning Strength=None2026.03—70.7
STTSModel=STTS, Pruning Strength=30%2026.03—79
STTSModel=STTS, Pruning Strength=40%2026.03—77.3
STTSModel=STTS, Pruning Strength=50%2026.03—77.7