| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 87.3 | Qwen3.5-122B-A10B | 2026-02-16 |
| Qwen3.5-Omni Technical Report | | 86.8 | Qwen3.5-Omni-Plus | 2026-04-21 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 86.7 | Qwen3.5-397B-A17B | 2026-02-16 |
| Qwen3.6 | ✓ Link | 86.6 | Qwen3.6-27B | 2026-04-21 |
| Qwen3.6 | ✓ Link | 86.2 | Qwen3.6-35B-A3B | 2026-04-21 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 85.9 | Qwen3.5-27B | 2026-02-16 |
| VideoJev: System One Model for Video Understanding and Interaction | ✓ Link | 85.9 | VideoJev-35B-A3B | 2026-10-13 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 85.6 | Qwen3.5-35B-A3B | 2026-02-16 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 84.4 | Qwen3.5-9B | 2026-02-16 |
| Qwen3-VL Technical Report | ✓ Link | 84.3 | Qwen3-VL-235B-A22B-Instruct | 2025-11-26 |
| Active Video Perception: Iterative Evidence Seeking for Agentic Long Video Understanding | | 84.3 | Active Video Perception (Gemini-2.5-Pro) | 2025-12-05 |
| META: Meta Evolution of Tool Trajectory Adaptation for Long-Video Understanding | | 83.6 | Qwen3-VL+META | 2026-06-03 |
| Kwai Keye-VL-2.0 Technical Report | ✓ Link | 82.8 | Keye-VL-2.0-30B-A3B | 2026-06-09 |
| HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning | | 82.52 | Qwen3.5-397B-A17B (RLVR w/ Multi-Hop) | 2026-03-17 |
| VideoJev: System One Model for Video Understanding and Interaction | ✓ Link | 80.6 | VideoJev-4B | 2026-10-13 |
| InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models | ✓ Link | 79.5 | InternVL3-78B | 2025-04-14 |
| Reasoning with Memory: A Temporal Granularity-Adaptive Framework for Training-Free Long Video Understanding | | 77.6 | ReMem (Qwen3-VL-8B-Instruct, 32 frames) | 2026-06-30 |
| InternVideo3: Agentify Foundation Models with Multimodal Contextual Reasoning | ✓ Link | 77.3 | InternVideo3-8B | 2026-06-10 |
| OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning | | 77.1 | OmniSeek (ours) | 2026-10-01 |
| AdaptToken: Entropy-based Adaptive Token Selection for MLLM Long Video Understanding | ✓ Link | 76.8 | Qwen2.5-VL + AdaptToken | 2026-03-30 |
| QCA: Query- and Content-Aware Keyframe Selection for Long Video Understanding | ✓ Link | 76.8 | QCA | 2026-07-01 |
| Divide, then Ground: Adapting Frame Selection to Query Types for Long-Form Video Understanding | ✓ Link | 76.66 | DIG (Qwen2.5-VL-32B) | 2025-12-03 |
| LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence | ✓ Link | 76.6 | LLaVA-OneVision-2-8B | 2026-05-25 |
| MiniCPM-o 4.5: Towards Real-Time Full-Duplex Omni-Modal Interaction | ✓ Link | 76.5 | MiniCPM-o 4.5-Instruct | 2026-04-30 |
| EventSTU: Event-Guided Efficient Spatio-Temporal Understanding for Video-LLMs | | 76.49 | EventSTU (Qwen3-VL-8B) | 2026-09-08 |
| Ovis2.5 Technical Report | ✓ Link | 76.4 | Ovis2.5-9B | 2025-08-15 |
| ReQuest: Rethinking-based Question-Aware Frame Selection for Long-Form Video QA | ✓ Link | 76.2 | Qwen3-VL + ReQuest | 2026-07-02 |
| Qwen3-Omni Technical Report | ✓ Link | 75.7 | Qwen3-Omni-Flash-Instruct | 2025-09-22 |
| Small Vision-Language Models are Smart Compressors for Long Video Understanding | ✓ Link | 75.6 | Tempo | 2026-04-09 |
| MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe | ✓ Link | 75.1 | MiniCPM-V 4.5 | 2025-09-16 |
| Qwen2.5-VL Technical Report | ✓ Link | 74.6 | Qwen2.5-VL-72B | 2025-02-19 |
| Rethinking Long-Video Efficiency: A Joint Allocation Perspective on Frames, Pixels, and Front-End Latency | | 74.38 | LoHi-SemDiv (Qwen3-VL-4B) | 2026-10-03 |
| Kimi-VL Technical Report | ✓ Link | 74.2 | Kimi-VL-A3B-Instruct | 2025-04-10 |
| Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding | ✓ Link | 74.0 | LatentStream (Qwen2.5-VL-7B) | 2026-09-03 |
| Zoom-Zero: Reinforced Coarse-to-Fine Video Understanding via Temporal Zoom-in | ✓ Link | 73.4 | Zoom-Zero + Divide-and-conquer | 2025-12-16 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 73.0 | VideoLLaMA3 7B | 2025-01-22 |
| OmniJigsaw: Enhancing Omni-Modal Reasoning via Modality-Orchestrated Reordering | ✓ Link | 72.63 | OmniJigsaw (SMS) | 2026-04-09 |
| Attend Before Attention: Efficient and Scalable Video Understanding via Autoregressive Gazing | ✓ Link | 71.6 | NVILA-8B-HD-Video | 2026-03-12 |
| Perceptual Projection Pruning: Diversity-Aware Video Token Pruning for Multimodal Large Language Models | | 71.1 | P3 (Qwen3-VL-32B, 25% retention) | 2026-09-08 |
| ResAdapt: Adaptive Resolution for Efficient Multimodal Reasoning | ✓ Link | 70.8 | Qwen3-VL-8B + ResAdapt (128 frames, 22.9% retention) | 2026-03-30 |
| LiteFrame: Efficient Vision Encoders Unlock Frame Scaling in Video LLMs | ✓ Link | 70.7 | LiteFrame | 2026-05-17 |
| VisReflect: Latent Visual Reflection for Fine-Grained Perception in Long Visual Context | ✓ Link | 70.6 | VisReflect-7B-Video | 2026-06-29 |
| Unified Spatio-Temporal Token Scoring for Efficient Video VLMs | ✓ Link | 70.5 | Molmo2 (SigLIP2 So400M + Qwen3-4B) with STTS (30%) + Test-Time Scaling (92 frames) | 2026-03-18 |
| Demo-ICL: In-Context Learning for Procedural Video Knowledge Acquisition | ✓ Link | 70.4 | Demo-ICL (7B) | 2026-02-09 |
| MeToM: Metadata-Guided Token Merging for Efficient Video LLMs | | 70.4 | LLaVA-Video-7B + MeToM | 2026-06-03 |
| Qwen2.5-VL Technical Report | ✓ Link | 70.2 | Qwen2.5-VL-7B | 2025-02-19 |
| FlexiVideo: Variation-Aware Temporal Dynamics Modeling for Efficient Video Understanding | ✓ Link | 69.6 | FlexiVideo-3B (extra data) | 2026-06-03 |
| From Pixels to Words -- Towards Native One-Vision Models at Scale | ✓ Link | 69.3 | NEO-ov 9B | 2026-05-27 |
| S3-Prune: Stability-Aware Token Budgeting for Long-Form Video-Language Models | | 68.89 | S3-prune (LLaVA-VID-7B, 25% vision-token retention) | 2026-09-08 |
| Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model | ✓ Link | 68.7 | Mage-VL-4B (tc32) | 2026-07-27 |
| EVA: Efficient Reinforcement Learning for End-to-End Video Agent | ✓ Link | 68.3 | EVA-GRPO | 2026-03-24 |
| LLaVA-OneVision: Easy Visual Task Transfer | ✓ Link | 68.0 | LLaVA-OneVision-72B | 2024-08-06 |
| CASA: Cross-Attention via Self-Attention for Efficient Vision-Language Fusion | ✓ Link | 67.9 | CAt+v (Qwen2.5-VL, 3B) | 2025-12-22 |
| FutureOmni: Evaluating Future Forecasting from Omni-Modal Context for Multimodal LLMs | ✓ Link | 67.86 | video-SALMONN 2 (7B) +OFF | 2026-01-20 |
| Training-free Uncertainty Guidance for Complex Visual Tasks with MLLMs | ✓ Link | 67.6 | UG-Sample (InternVL2.5-8B) | 2025-10-01 |
| Swift Sampling: Selecting Temporal Surprises via Taylor Series | | 67.2 | Swift Sampling (LLaVA-Video) | 2026-05-21 |
| Rethinking Chain-of-Thought Reasoning for Videos | ✓ Link | 67.0 | Qwen2.5-VL-7B (GRPO, concise reason, trained with token compression & 6x frames) | 2025-12-10 |
| LENS: Adaptive Spatio-Temporal Zooming for Keyframe Sampling in Long-Form Videos | ✓ Link | 66.7 | LENS (Qwen2.5-VL 7B, 32 frames) | 2026-07-27 |
| TimeViper: A Hybrid Mamba-Transformer Vision-Language Model for Efficient Long Video Understanding | ✓ Link | 65.6 | TimeViper | 2025-11-20 |
| LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding | ✓ Link | 65.4 | LongVU | 2024-10-22 |
| H2SVC: Head-aware Heterogeneous Streaming Video Cache for Online Video Understanding | | 65.1 | LLaVA-OV-7B + H2SVC | 2026-09-08 |
| Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation | ✓ Link | 64.9 | Intern Lumina U2 (NVIDIA) | 2026-10-14 |
| Incentivizing Versatile Video Reasoning in MLLMs via Data-Efficient Reinforcement Learning | ✓ Link | 64.6 | VideoReasoner | 2026-06-03 |
| VideoDetective: Clue Hunting via both Extrinsic Query and Intrinsic Relevance for Long Video Understanding | ✓ Link | 63.8 | VideoDetective (SeedVL-1.5, 20B) | 2026-03-23 |
| Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone | ✓ Link | 61.1 | Dream-VL | 2025-12-27 |
| LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning | ✓ Link | 59.5 | LLaDA-V | 2025-05-22 |
| SmolVLM: Redefining small and efficient multimodal models | ✓ Link | 55.2 | SmolVLM 2.2B | 2025-04-07 |
| Less Data, Faster Convergence: Goal-Driven Data Optimization for Multimodal Instruction Tuning | ✓ Link | 46.89 | GDO (Temp+) | 2026-03-12 |
| VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs | ✓ Link | 45.6 | VideoLLaMA 2 72B | 2024-06-11 |