Long-Form Video Understanding on LVBench
94.4Overall ScoreHuman Expert
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Human ExpertReason Model=Human2026.06 | 94.4 | — | — | — | — | — | — | |
| MEMDREAMERReason Model=Gemini-3.1-Pro2026.06 | 90.7 | 90.1 | 90.6 | 91.4 | 91.8 | 89.6 | 89.7 | |
| MEMDREAMERReason Model=Qwen3-VL-235B-A22B-Thinking2026.06 | 84.8 | 84.6 | 85.2 | 85.6 | 87.7 | 80.6 | 84.5 | |
| MEMDREAMERReason Model=Gemini-2.5-Pro2026.06 | 80.7 | 77.1 | 83.6 | 82.5 | 86.4 | 81.1 | 91.4 | |
| VideoARMReason Model=OpenAI-o32026.06 | 79.7 | — | — | — | — | — | — | |
| Gemini-3.1-ProReason Model=Gemini-3.1-Pro2026.06 | 78.2 | 78.1 | 76.7 | 86.3 | 81.4 | 74.1 | 70.7 | |
| DVDReason Model=OpenAI-o32026.06 | 74.2 | 73.4 | 73.3 | 80.4 | 72.3 | 70.7 | 74.1 | |
| Deep Video Discovery2026.07 | 74.2 | 73.4 | 73.3 | 80.4 | 72.3 | 70.6 | 74.1 | |
| Gemini-2.5-ProReason Model=Gemini-2.5-Pro2026.06 | 72 | 71.5 | 71.1 | 80 | 69.1 | 67.7 | 63.5 | |
| Qwen3.5Reason Model=Qwen3.5-35A3B2026.06 | 71.3 | 72.2 | 69.4 | 68.9 | 66.2 | 74.4 | 67.5 | |
| VideoSeekReason Model=GPT-52026.06 | 68.4 | — | — | — | — | — | — | |
| Seed1.5VLReason Model=Seed1.5VL-Thinking2026.06 | 64.6 | 65.4 | 63.4 | 53.6 | 46.6 | 68 | 63.7 | |
| Seed1.5-VL-ThinkingLLM=200B2026.07 | 64.6 | 65.4 | 63.4 | 68 | 53.6 | 63.7 | 46.6 | |
| Qwen3-VLReason Model=Qwen3-VL-235B-A22B-Thinking2026.06 | 63.6 | 63.7 | 62.6 | 62.6 | 59.6 | 63.1 | 65.5 | |
| WorldMM-GPTReason Model=GPT-52026.06 | 61.9 | — | — | — | — | — | — | |
| MR. VideoReason Model=MR. Video2026.06 | 60.8 | 59.8 | 57.4 | 71.4 | 58.8 | 57.7 | 50 | |
| GLM-4.6VReason Model=GLM-4.6V-106B-A12B2026.06 | 59.5 | 58.8 | 59.2 | 57.8 | 53.9 | 69.7 | 61.8 | |
| OpenAI-o3Reason Model=OpenAI-o32026.06 | 57.1 | 57.6 | 56.4 | 62.9 | 46.8 | 50.8 | 67.2 | |
| GLM-4.5VReason Model=GLM-4.5V-106B-A12B2026.06 | 53.4 | 55.3 | 53.1 | 55.3 | 49.9 | 57.5 | 39.6 | |
| AdaRETAKEReason Model=AdaRETAKE2026.06 | 53.3 | 53 | 50.7 | 62.2 | 45.5 | 54.7 | 37.9 | |
| AdaReTaKeLLM=72B, Frames=≤10242026.07 | 53.3 | 53 | 50.7 | 62.2 | 45.5 | 54.7 | 37.9 | |
| LLaVA-Video-7B† + Ours (Query-Conditioned Evidential Keyframe Sampling)Frames=642026.04 | 49.4 | 52.4 | 45.9 | 54.6 | 39 | 46.8 | 32.8 | |
| M3-AgentReason Model=M3-Agent2026.06 | 49.3 | — | — | — | — | — | — | |
| QSVideoLLM=8B, Frames=322026.07 | 49.1 | 51.1 | 45.6 | 54.6 | 39.6 | 53.2 | 32.8 | |
| GPT-4oReason Model=GPT-4o2026.06 | 48.9 | 48.9 | 49.5 | 48.1 | 40.9 | 50.3 | 50 | |
| GPT-4oFrames=602026.07 | 48.9 | 48.9 | 49.5 | 48.1 | 40.9 | 50.3 | 50 | |
| GLM-4V-PlusFrames=≤3002026.07 | 48.7 | 46.2 | 47.8 | 54.1 | 42.7 | 46.5 | 37.9 | |
| Gemini-2.0-FlashReason Model=Gemini-2.0-Flash2026.06 | 48.6 | 47.4 | 48.5 | 56.8 | 39.3 | 44.4 | 41.4 | |
| LLaVA-Video-7B† + ReTaKeFrames=≤ 10242026.04 | 48.5 | — | — | — | — | — | — | |
| AdaSparkModel Size=Mid Size, Backbone=Qwen-2.5-VL-7B, SFT=true2026.04 | 47.9 | — | — | — | — | — | — | |
| Qwen2-VL-7B† + ReTaKeFrames=≤ 20482026.04 | 47.8 | — | — | — | — | — | — | |
| Qwen2.5-VL-7B† + Ours (Query-Conditioned Evidential Keyframe Sampling)Frames=322026.04 | 47.7 | 49.8 | 44.5 | 57 | 37.3 | 41.8 | 34.5 | |
| Video-o3Frames=≤ 7682026.04 | 47.6 | — | — | — | — | — | — | |
| Qwen2-VL-7B† + Ours (Query-Conditioned Evidential Keyframe Sampling)Frames=322026.04 | 46.6 | 47.9 | 44.5 | 55 | 42.7 | 43.8 | 25.9 | |
| Qwen2-VL-7B† + TSPOFrames=642026.04 | 46.4 | — | — | — | — | — | — | |
| SEALLLM=34B, Frames=162026.07 | 45.9 | 47.9 | 41.3 | 51.5 | 32.3 | 43.3 | 39.7 | |
| QSVideoLLM=8B, Frames=162026.07 | 45.8 | 47.3 | 41.7 | 49.8 | 35.5 | 50.8 | 32.8 | |
| VideoLLaMA3-7BFrames=-2026.04 | 45.3 | — | — | — | — | — | — | |
| LLaVA-Video-7B† + TSPOFrames=642026.04 | 45.3 | — | — | — | — | — | — | |
| Qwen-2.5-VL-7BModel Size=Mid Size, SFT=false2026.04 | 45.3 | — | — | — | — | — | — | |
| Qwen-2.5-VL-7B + SFTModel Size=Mid Size, SFT=true2026.04 | 45.2 | — | — | — | — | — | — | |
| AdaSparkModel Size=Small Size, Backbone=Qwen-2.5-VL-3B, SFT=true2026.04 | 45 | — | — | — | — | — | — | |
| InternVL3.5Reason Model=InternVL3.5-30BA3B2026.06 | 44.4 | 42.7 | 44.1 | 46.4 | 40.9 | 48.3 | 36.2 | |
| InternVL2.5Reason Model=InternVL2.5-78B2026.06 | 43.6 | 43.8 | 42 | 37.9 | 36.8 | 51 | 42.1 | |
| InternVL2.5-78BLLM=72B, Frames=162026.07 | 43.6 | 43.8 | 42 | 42.1 | 36.8 | 51 | 37.9 | |
| FrameFusionModel Size=Small Size, Backbone=Qwen-2.5-VL-3B, SFT=true2026.04 | 43.4 | — | — | — | — | — | — | |
| Qwen-2.5-VL-3BModel Size=Small Size, SFT=false2026.04 | 43.3 | — | — | — | — | — | — | |
| FastVModel Size=Small Size, Backbone=Qwen-2.5-VL-3B, SFT=true2026.04 | 43.3 | — | — | — | — | — | — | |
| FastVModel Size=Mid Size, Backbone=Qwen-2.5-VL-7B, SFT=true2026.04 | 43.3 | — | — | — | — | — | — | |
| Qwen-2.5-VL-3B + SFTModel Size=Small Size, SFT=true2026.04 | 43.1 | — | — | — | — | — | — | |
| ToMeModel Size=Mid Size, Backbone=Qwen-2.5-VL-7B, SFT=true2026.04 | 43.1 | — | — | — | — | — | — | |
| VideoChat-Flash-2BModel Size=Small Size2026.04 | 42.9 | — | — | — | — | — | — | |
| FrameFusionModel Size=Mid Size, Backbone=Qwen-2.5-VL-7B, SFT=true2026.04 | 42.8 | — | — | — | — | — | — | |
| ToMeModel Size=Small Size, Backbone=Qwen-2.5-VL-3B, SFT=true2026.04 | 42.4 | — | — | — | — | — | — | |
| MoBAModel Size=Mid Size, Backbone=Qwen-2.5-VL-7B, SFT=true2026.04 | 42.3 | — | — | — | — | — | — | |
| MiniCPM-o 2.6LLM=8B, Frames=642026.07 | 42.3 | 43 | 42.5 | 39.9 | 39.1 | 46.8 | 31 | |
| MiniCPM-o 2.6LLM=8B, Frames=322026.07 | 42 | 41.4 | 40.7 | 39.2 | 35.5 | 48.3 | 37.9 | |
| LLaVA-Video-7B†Frames=642026.04 | 41.7 | 41.5 | 40.2 | 42.3 | 33.2 | 49.8 | 29.3 | |
| VCAFrames=-2026.04 | 41.3 | 43.7 | 40.7 | 37.8 | 38 | 46.2 | 27.3 | |
| TimeMarker-8BFrames=≤1282026.04 | 41.3 | 42.8 | 39.1 | 34.9 | 38.7 | 38.2 | 48.8 | |
| VCAReason Model=VCA2026.06 | 41.3 | 43.7 | 40.7 | 37.8 | 38 | 46.2 | 27.3 | |
| Qwen2-VL-72BLLM=72B, Frames=482026.07 | 41.3 | 38 | 41.1 | 38.3 | 41.4 | 46.5 | 46.6 | |
| TimeMarkerLLM=8B, Frames=≤1282026.07 | 41.3 | 42.8 | 39.1 | 34.9 | 38.7 | 38.2 | 48.8 | |
| InternVL2-40BLLM=34B, Frames=162026.07 | 39.6 | 37.4 | 39.7 | 43.4 | 31.4 | 42.5 | 41.4 | |
| MoBAModel Size=Small Size, Backbone=Qwen-2.5-VL-3B, SFT=true2026.04 | 39.4 | — | — | — | — | — | — | |
| Qwen2-VL-7B†Frames=322026.04 | 39.1 | 38.7 | 39 | 36.8 | 37.3 | 39.8 | 29.3 | |
| MiniCPM-o 2.6LLM=8B, Frames=162026.07 | 38.9 | 37.1 | 38.2 | 37.5 | 32.7 | 46.3 | 41.4 | |
| Video-Thinker-7BFrames=322026.04 | 38.4 | — | — | — | — | — | — | |
| GLM-4V-PlusFrames=302026.07 | 38.3 | 39.9 | 35.8 | 34.8 | 37.7 | 40 | 32.8 | |
| Qwen2.5-VL-7B†Frames=322026.04 | 37.6 | 36.8 | 38.6 | 40.6 | 32.7 | 37.3 | 29.3 | |
| Video-R1-7BFrames=322026.04 | 37.4 | — | — | — | — | — | — | |
| Qwen2.5-VL-7B† + FrameThinkerFrames=23.92026.04 | 36.6 | — | — | — | — | — | — | |
| VideoAgentFrames=-2026.04 | 29.3 | 28 | 30.3 | 28 | 29.3 | 28 | 36.4 | |
| VideoAgentReason Model=VideoAgent2026.06 | 29.3 | 28 | 30.3 | 28 | 29.3 | 28 | 36.4 | |
| VideoTreeFrames=-2026.04 | 28.8 | 30.3 | 25.1 | 26.5 | 27.7 | 31.9 | 25.5 | |
| VideoTreeReason Model=VideoTree2026.06 | 28.8 | 30.3 | 25.1 | 26.5 | 27.7 | 31.9 | 25.5 | |
| MovieChat-7BFrames=>100002026.04 | 22.5 | 21.3 | 23.1 | 25.9 | 22.3 | 24 | 17.2 |