List your company
ResearchFormerly WizWand

General Video Understanding on Video-MME (Accuracy)

87.8AccuracyGemini 1.5 Pro
53.27262.23671.280.164May 18, 2026May 26, 2026Jun 3, 2026Jun 11, 2026Jun 19, 2026Jun 27, 2026Jul 6, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Gemini 1.5 ProSource=[60]2026.0687.8
Gemini 1.5 ProSource=[59]2026.0687.5
GPT-4o2026.0686.9
Gemini 1.5 Flash2026.0684.2
GPT-4o mini2026.0682.3
Claude 3.5 Sonnet2026.0680.5
SpatialClawBackbone=Gemma4-31B2026.0677
Key-VL-1.5-8B2026.0676.2
Molmo2-8B2026.0675.8
Eagle-X5-7B2026.0675.7
GLM-4V-9B2026.0675.6
AdaCodecToken budget=comparable, Backbone=Qwen2-VL-7B2026.0675.5
Qwen2-VL-7B2026.0675.2
Gemini-1.5-ProFrames=322026.0575
AdaCodecToken budget=1/7, Backbone=Qwen2-VL-7B2026.0675
Gemini-1.5-ProSize=-, Zero-shot=true, Category=Proprietary Models2026.0775
SpaceTools ToolshedBackbone=Gemma4-31B2026.0674.6
Gemini-3.1-ProModel Type=Proprietary2026.0574.3
MiniCPM-V-2.6-8B2026.0673.5
SELECTSTREAM-Qwen3-VL-8BSize=8B, #Frames=1 fps, max 10242026.0673.2
GPT-5.1Model Type=Proprietary2026.0572.9
GPT-4oFrames=322026.0571.9
GPT-4oSize=-, Zero-shot=true, Category=Proprietary Models2026.0771.9
Qwen3-VL-8BSize=8B, #Frames=2 fps, max 20482026.0671.4
LLaVA-Video-7B2026.0669.7
VideoChat-Flash-7B2026.0669.7
CoPE-VideoLM-7B2026.0669.4
Molmo2-O-7B2026.0669.2
InternVL2-8B2026.0668.6
Streamo-7BSize=7B, #Frames=1 fps2026.0667.9
SELECTSTREAM-Qwen2.5-VL-7BSize=7B, #Frames=1 fps, max 10242026.0667.8
Qwen-2.5-VLSize=7B, Zero-shot=true, Category=Supervised Fine-Tuning Models2026.0766.1
InternVL3-8BModel Type=Open-Source2026.0565.6
Qwen3-VL-8B + CRPOBackbone=Qwen3-VL-8B, Post-training Method=CRPO2026.0565.6
TimeThinkSize=7B, Zero-shot=true, Category=Reinforcement Fine-Tuning Models2026.0765.5
PLM-7B2026.0665.4
VideoChat-R1.5Size=7B, Zero-shot=true, Category=Reinforcement Fine-Tuning Models2026.0765.2
Qwen3-VL-8B + ArrowRLBackbone=Qwen3-VL-8B, Post-training Method=ArrowRL2026.0565.1
Qwen2.5-VL-7BSize=7B, #Frames=max 7682026.0665.1
Qwen3-VL-8B + T-GRPOBackbone=Qwen3-VL-8B, Post-training Method=T-GRPO2026.0565
Qwen3-VL-8BBackbone=Qwen3-VL-8B, Post-training Method=base2026.0564.9
Qwen3-VL-8B + GRPOBackbone=Qwen3-VL-8B, Post-training Method=GRPO2026.0564.5
ReMoRa-7B2026.0664.4
Qwen2.5-VL-GRPOSize=7B, Zero-shot=true, Category=Reinforcement Fine-Tuning Models2026.0764.3
InternVL2.5Size=8B, Zero-shot=true, Category=Supervised Fine-Tuning Models2026.0764.2
VideoChat-R1Size=7B, Zero-shot=true, Category=Reinforcement Fine-Tuning Models2026.0764.1
VideoLatent-7B (Ours)Frames=64, Architecture Category=Latent MLLM2026.0663.8
Qwen2.5-VL-SFTSize=7B, Zero-shot=true, Category=Supervised Fine-Tuning Models2026.0763.6
Qwen3-VL-4B + T-GRPOBackbone=Qwen3-VL-4B, Post-training Method=T-GRPO2026.0563.4
Qwen2-VL-7BSize=7B, #Frames=642026.0663.3
LLaVA-Video-7BSize=7B, #Frames=642026.0663.3
Qwen3-VL-4B + CRPOBackbone=Qwen3-VL-4B, Post-training Method=CRPO2026.0563
Qwen2.5-VL-7BFrames=64, Architecture Category=Standard MLLM, CoT=false2026.0662.7
Qwen3-VL-4BBackbone=Qwen3-VL-4B, Post-training Method=base2026.0562.2
Qwen3-VL-4B + ArrowRLBackbone=Qwen3-VL-4B, Post-training Method=ArrowRL2026.0562.2
Qwen3-VL-4B + GRPOBackbone=Qwen3-VL-4B, Post-training Method=GRPO2026.0562.1
Mull-Token-7BFrames=64, Architecture Category=Latent MLLM2026.0662.1
LVR-7BFrames=64, Architecture Category=Latent MLLM2026.0662
StreamForest-7B (FT-drive)Size=7B, #Frames=1 fps2026.0661.9
StreamForest-7BSize=7B, #Frames=1 fps2026.0661.4
VideoLatent-7B (Ours)Frames=32, Architecture Category=Latent MLLM2026.0661.4
Video-R1-7BFrames=64, Architecture Category=Standard MLLM2026.0661.4
Open-o3-Video-7BFrames=64, Architecture Category=Standard MLLM2026.0661.4
Video-R1Size=7B, Zero-shot=true, Category=Reinforcement Fine-Tuning Models2026.0761.4
Open-o3-Video-7BFrames=32, Architecture Category=Standard MLLM2026.0661.3
STORMFrames=322026.0561
Qwen2.5-VL-7B-Instruct(CoT)Frames=322026.0560.8
LongVU-7BSize=7B, #Frames=1 fps2026.0660.6
LVR-7BFrames=32, Architecture Category=Latent MLLM2026.0660.5
Qwen2.5-VL-7BModel Type=Open-Source2026.0560.3
LLaVA-Next-VideoFrames=322026.0560.2
ArrowRL*Model Type=Open-Source2026.0560.1
VILA-1.5Size=40B, Zero-shot=true, Category=Supervised Fine-Tuning Models2026.0760.1
Video-Thinker-7BFrames=32, method_variant=think with video2026.0560
VideoRFT-7BFrames=32, Architecture Category=Standard MLLM2026.0659.8
VideoRFTSize=7B, Zero-shot=true, Category=Reinforcement Fine-Tuning Models2026.0759.8
LongVT-RFTFrames=32, method_variant=think with tools2026.0559.7
Qwen2.5-VL-7B (CoT)Frames=64, Architecture Category=Standard MLLM, CoT=true2026.0659.6
Mull-Token-7BFrames=32, Architecture Category=Latent MLLM2026.0659.5
GPT-4VSize=-, Zero-shot=true, Category=Proprietary Models2026.0759.5
Qwen2.5-VL-7BFrames=32, Architecture Category=Standard MLLM, CoT=false2026.0659.4
VideoLatent-7B (Ours)Frames=16, Architecture Category=Latent MLLM2026.0659.3
Video-R1-7BFrames=32, Architecture Category=Standard MLLM2026.0659.3
LVR-7BFrames=16, Architecture Category=Latent MLLM2026.0659.1
LLaVA-OneVision-7BSize=7B, #Frames=322026.0658.2
LLaVA-OVSize=7B, Zero-shot=true, Category=Supervised Fine-Tuning Models2026.0758.2
LLaVA-OV-7BModel Type=Open-Source2026.0557.6
Video-R1-7BFrames=322026.0557.4
Video-R1-7BFrames=16, Architecture Category=Standard MLLM2026.0657.4
Dispider-7BSize=7B, #Frames=1 fps2026.0657.2
Mull-Token-7BFrames=16, Architecture Category=Latent MLLM2026.0657
Qwen2.5-VL-7BFrames=16, Architecture Category=Standard MLLM, CoT=false2026.0656.6
Qwen2.5-VL-7B (CoT)Frames=32, Architecture Category=Standard MLLM, CoT=true2026.0656.6
KangarooFrames=322026.0556
VideoRefer2026.0555.9
SWIM2026.0555.9
LLaVA-Octopus2026.0555.7
Qwen2.5-VL-7B-SFTFrames=322026.0555.4
VideoLLaMA2.12026.0554.9
VideoChat2-HD2026.0554.6