Long-video Question Answering on Long VideoBench (val)
73.2AccuracyGOPAgen
Evaluation Results
| Method | Links | |
|---|---|---|
| GOPAgenType=Agentic video understanding framework2026.06 | 73.2 | |
| GPT-5LLM=N/A, # Frames=N/A2025.12 | 72.6 | |
| Deep Video DiscoveryType=Agentic video understanding framework2026.06 | 71.6 | |
| VideoLucyType=Agentic video understanding framework2026.06 | 70.2 | |
| OpenAI o3Access=Closed-Source2026.06 | 67.5 | |
| AdaReTaKeAccess=Open-Source2026.06 | 67 | |
| GPT-4o2025.01 | 66.7 | |
| GPT-4oFrames=2562025.02 | 66.7 | |
| GPT-4oLLM=N/A, # Frames=N/A2025.12 | 66.7 | |
| GPT-4oAccess=Closed-Source2026.06 | 66.7 | |
| Gemini-1.5-Pro2025.01 | 64 | |
| Gemini-1.5-ProFrames=2562025.02 | 64 | |
| Gemini-1.5-ProLLM=N/A, # Frames=N/A2025.12 | 64 | |
| Gemini-1.5-ProAccess=Closed-Source2026.06 | 64 | |
| InternVL2.5-78B2025.01 | 63.6 | |
| InternVL2.5-78BAccess=Open-Source2026.06 | 63.6 | |
| LLaVA-Video + OneClip-RAGLLM=7B, # Frames=642025.12 | 62.9 | |
| LLaVA-Video w/ AKSFrames=64, LLM=7B, sampling=AKS2025.02 | 62.7 | |
| AKSLLM=7B, # Frames=642025.12 | 62.7 | |
| InternVL3.5LLM=8B, # Frames=642025.12 | 62.1 | |
| LLaVA-Video-72B2025.01 | 61.9 | |
| Gemini-1.5-FlashFrames=2562025.02 | 61.6 | |
| GPT-4VFrames=2562025.02 | 61.3 | |
| Qwen2.5-VL-72BAccess=Open-Source2026.06 | 60.7 | |
| Qwen2-VL w/ AKSFrames=32, LLM=7B, sampling=AKS2025.02 | 60.5 | |
| InternVL2.5-8B2025.01 | 60 | |
| VideoNSALLM=7B, # Frames=20482025.12 | 60 | |
| mPLUG-Owl3Access=Open-Source2026.06 | 59.8 | |
| LLaVA-OV w/ AKSFrames=32, LLM=7B, sampling=AKS2025.02 | 59.3 | |
| LLaVA-VideoLLM=7B, # Frames=642025.12 | 59.3 | |
| LLaVA-VideoFrames=64, LLM=7B2025.02 | 58.9 | |
| Tarsier2-7BNumber of frames=128f2025.01 | 58.6 | |
| LLaVA-Video-7B2025.01 | 58.2 | |
| NVILALLM=8B, # Frames=2562025.12 | 57.7 | |
| Qwen2.5-VLLLM=7B, # Frames=322025.12 | 56 | |
| Qwen2-VL-7B2025.01 | 55.6 | |
| Qwen2-VLFrames=32, LLM=7B2025.02 | 55.5 | |
| MiniCPM-V 2.6Frames=64, LLM=8B2025.02 | 54.9 | |
| LLaVA-OVFrames=32, LLM=7B2025.02 | 54.8 | |
| Tarsier-34B2025.01 | 54.2 | |
| PLLAVAFrames=32, LLM=34B2025.02 | 53.2 | |
| mPLUG-Owl3LLM=7B, # Frames=162025.12 | 52.1 | |
| Video-XLLLM=7B, # Frames=1282025.12 | 50.7 | |
| LongVALLM=7B, # Frames=2562025.12 | 47.8 | |
| Tarsier-7B2025.01 | 39.8 | |
| VideoLLaVAFrames=8, LLM=7B2025.02 | 39.1 |