| Kimi-VL Technical Report | ✓ Link | 78.5 | Kimi-VL-A3B-Instruct | 2025-04-10 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 77.9 | Qwen2-VL-72B | 2024-09-18 |
| InternVideo3: Agentify Foundation Models with Multimodal Contextual Reasoning | ✓ Link | 76.6 | InternVideo3-8B | 2026-06-10 |
| Qwen2.5-VL Technical Report | ✓ Link | 76.2 | Qwen2.5-VL-72B | 2025-02-19 |
| Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context | ✓ Link | 72.2 | Gemini 1.5 Pro | 2024-03-08 |
| Hello GPT-4o | | 72.2 | GPT-4o | 2024-05-13 |
| Introducing Gemini 2.0: our new AI model for the agentic era | | 71.5 | Gemini 2.0 Flash Experimental | 2024-12-11 |
| BIMBA: Selective-Scan Compression for Long-Range Video Question Answering | ✓ Link | 71.14 | BIMBA-LLaVA-Qwen2-7B | 2025-03-12 |
| Penguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders | ✓ Link | 70.2 | Penguin-VL 8B | 2026-03-06 |
| LinVT: Empower Your Image-level Large Language Model to Understand Videos | ✓ Link | 69.5 | LinVT-Qwen2-VL(7B) | 2024-12-06 |
| Progressive Video Condensation with MLLM Agent for Long-form Video Understanding | | 69.3 | ProVCA | 2026-04-03 |
| PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding | ✓ Link | 68.8 | PLM-8B | 2025-04-17 |
| Qwen2.5-Omni Technical Report | ✓ Link | 68.6 | Qwen2.5-Omni | 2025-03-26 |
| LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding | ✓ Link | 67.6 | LongVU (7B) | 2024-10-22 |
| Attend Before Attention: Efficient and Scalable Video Understanding via Autoregressive Gazing | ✓ Link | 66.9 | NVILA-8B-HD-Video | 2026-03-12 |
| Aria: An Open Multimodal Native Mixture-of-Experts Model | ✓ Link | 66.8 | Aria | 2024-10-08 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 66.7 | Qwen2-VL-7B | 2024-09-18 |
| Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension | ✓ Link | 66.7 | Video-RAG (Based on LLaVA-Video) | 2024-11-20 |
| LENS: Adaptive Spatio-Temporal Zooming for Keyframe Sampling in Long-Form Videos | ✓ Link | 66.1 | LENS (Qwen2.5-VL 7B, 32 frames) | 2026-07-27 |
| HiCrew: Hierarchical Reasoning for Long-Form Video Understanding via Question-Aware Multi-Agent Collaboration | | 64.6 | HiCrew | 2026-04-23 |
| CurveStream: Boosting Streaming Video Understanding in MLLMs via Curvature-Aware Hierarchical Visual Memory Management | ✓ Link | 64.29 | Qwen2.5-VL-7B (with CurveStream) | 2026-03-20 |
| VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs | ✓ Link | 63.9 | VideoLLaMA2 (72B) | 2024-06-11 |
| Bridging VideoQA and Video-Guided Agentic Tasks via Generalized Keyframe Extraction | ✓ Link | 63.6 | TASKER (GPT-4o) | 2026-06-28 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 63.3 | VideoLLaMA3 7B | 2025-01-22 |
| Bridging VideoQA and Video-Guided Agentic Tasks via Generalized Keyframe Extraction | ✓ Link | 63.1 | TASKER (GPT-4) | 2026-06-28 |
| MeToM: Metadata-Guided Token Merging for Efficient Video LLMs | | 62.4 | LLaVA-OneVision-7B + MeToM | 2026-06-03 |
| H2SVC: Head-aware Heterogeneous Streaming Video Cache for Online Video Understanding | | 62.1 | LLaVA-OV-7B + H2SVC | 2026-09-08 |
| LLaVA-OneVision: Easy Visual Task Transfer | ✓ Link | 62.0 | LLaVA-OneVision-72B | 2024-08-06 |
| Perception Encoder: The best visual embeddings are not at the output of the network | ✓ Link | 62.0 | PE-Lang G | 2025-04-17 |
| Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding | ✓ Link | 62.0 | Molmo2-8B | 2026-01-15 |
| Tarsier: Recipes for Training and Evaluating Large Video Description Models | ✓ Link | 61.7 | Tarsier (34B) | 2024-06-30 |
| Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models | ✓ Link | 61.7 | AOT (LLaVA-OneVision-7B) | 2026-03-02 |
| S3-Prune: Stability-Aware Token Budgeting for Long-Form Video-Language Models | | 61.11 | S3-prune (LLaVA-OV-7B, 25% vision-token retention) | 2026-09-08 |
| VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videos | ✓ Link | 61.1 | VideoTree (GPT4) | 2024-05-29 |
| Too Many Frames, Not All Useful: Efficient Strategies for Long-Form Video QA | ✓ Link | 61.1 | LVNet | 2024-06-13 |
| OmniStream: Mastering Perception, Reconstruction and Action in Continuous Streams | ✓ Link | 60.9 | OmniStream-7B | 2026-03-12 |
| Accelerating Streaming Video Large Language Models via Hierarchical Token Compression | ✓ Link | 60.8 | ReKV with STC-Pruner | 2025-11-30 |
| EarlyTom: Early Token Compression Completes Fast Video Understanding | ✓ Link | 60.6 | EarlyTom (LLaVA-OneVision-7B) | 2026-05-28 |
| HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding | ✓ Link | 60.29 | HERMES (LLaVA-OV-7B, 4K tokens) | 2026-01-21 |
| InternVideo2: Scaling Foundation Models for Multimodal Video Understanding | ✓ Link | 60.2 | InternVideo2-6B | 2024-03-22 |
| TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning | ✓ Link | 60.0 | VideoChat-T (7B) | 2024-10-25 |
| MVBench: A Comprehensive Multi-modal Video Understanding Benchmark | ✓ Link | 56.7 | VideoChat2_phi3 | 2023-11-28 |
| MVBench: A Comprehensive Multi-modal Video Understanding Benchmark | ✓ Link | 55.8 | VideoChat2_HD_mistral | 2023-11-28 |
| Penguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders | ✓ Link | 55.7 | Penguin-VL 2B | 2026-03-06 |
| Flexible Frame Selection for Efficient Video Reasoning | | 55.6 | OpenLLoVi (Gemma) + FFS (Ours; @16.8→[4,32]) | 2025-06-11 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 54.9 | Qwen2-VL-2B | 2024-09-18 |
| MVBench: A Comprehensive Multi-modal Video Understanding Benchmark | ✓ Link | 54.4 | VideoChat2_mistral | 2023-11-28 |
| Vamos: Versatile Action Models for Video Understanding | ✓ Link | 53.6 | Vamos (GPT-4o) | 2023-11-22 |
| TraveLER: A Modular Multi-LMM Agent Framework for Video Question-Answering | ✓ Link | 53.3 | TraveLER | 2024-04-01 |
| A Simple LLM Framework for Long-Range Video Question-Answering | ✓ Link | 50.3 | LLoVi (GPT-3.5) | 2023-12-28 |
| Video ReCap: Recursive Captioning of Hour-Long Videos | ✓ Link | 50.23 | Video ReCap | 2024-02-20 |
| Vamos: Versatile Action Models for Video Understanding | ✓ Link | 48.3 | Vamos (GPT-4) | 2023-11-22 |
| Language Repository for Long Video Understanding | ✓ Link | 41.2 | LangRepo (12B) | 2024-03-21 |
| Video Mamba Suite: State Space Model as a Versatile Alternative for Video Understanding | ✓ Link | 38.70 | TimeMamba | 2024-03-14 |
| Understanding Long Videos with Multimodal Language Models | ✓ Link | 37.6 | MVU (13B) | 2024-03-25 |
| Vamos: Versatile Action Models for Video Understanding | ✓ Link | 36.7 | Vamos (13B) | 2023-11-22 |
| A Simple LLM Framework for Long-Range Video Question-Answering | ✓ Link | 33.5 | LLoVi (7B) | 2023-12-28 |
| TimeChat: A Time-sensitive Multimodal Large Language Model for Long Video Understanding | ✓ Link | 33.0 | TimeChat (7B) | 2023-12-04 |
| InternVideo: General Video Foundation Models via Generative and Discriminative Learning | ✓ Link | 32.1 | InternVideo | 2022-12-06 |
| mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality | ✓ Link | 31.1 | mPLUG-Owl (7B) | 2023-04-27 |
| Zero-Shot Video Question Answering via Frozen Bidirectional Language Models | ✓ Link | 26.9 | FrozenBiLM | 2022-06-16 |
| Self-Chained Image-Language Model for Video Localization and Question Answering | ✓ Link | 22.7 | SeViLA (4B) | 2023-05-11 |
| | 20.0 | Random | |