| Qwen3.8-Max: A New Bar for Coding and Cowork | ✓ Link | 85.9 | Qwen3.8-Max | 2026-08-03 |
| ReaGEN: Adaptive Generation of Structured Chains-of-Thought for Efficient Multimodal Reasoning | ✓ Link | 84.49 | ReaGEN (Qwen3-VL-4B, 4 iter) | 2026-06-03 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 83.8 | Qwen3.5-397B-A17B | 2026-02-16 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 82.9 | Qwen3.5-122B-A10B | 2026-02-16 |
| Masking Teacher and Reinforcing Student for Distilling Vision-Language Models | | 82.0 | InternVL3 - Masters | 2025-12-23 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 81.9 | Qwen3.5-35B-A3B | 2026-02-16 |
| Qwen3.6 | ✓ Link | 81.4 | Qwen3.6-27B | 2026-04-21 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 81.0 | Qwen3.5-27B | 2026-02-16 |
| SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture | ✓ Link | 80.92 | SenseNova-U1-A3B-MoT | 2026-05-12 |
| Qwen3.6 | ✓ Link | 80.7 | Qwen3.6-35B-A3B | 2026-04-21 |
| HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning | | 80.67 | Qwen3.5-397B-A17B (RLVR w/ Multi-Hop) | 2026-03-17 |
| Kwai Keye-VL 1.5 Technical Report | ✓ Link | 80.5 | Keye-VL-1.5-8B | 2025-09-01 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 79.7 | Qwen3.5-9B | 2026-02-16 |
| Qwen3.5-Omni Technical Report | | 79.4 | Qwen3.5-Omni-Plus | 2026-04-21 |
| TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models | ✓ Link | 79.3 | Tuna | 2025-12-01 |
| Qwen3-VL Technical Report | ✓ Link | 78.7 | Qwen3-VL-235B-A22B-Thinking | 2025-11-26 |
| SpatialASM: Internalizing Active Mentalization for Spatial Reasoning | ✓ Link | 78.5 | SpatialASM-Qwen3.5-9B | 2026-10-09 |
| STEP3-VL-10B Technical Report | ✓ Link | 77.64 | Step3-VL-10B (PaCoRe) | 2026-01-14 |
| SenseNova-U1.5: Towards Native Unified Visual Intelligence | ✓ Link | 77.53 | SenseNova-U1.5 8B-Think | 2026-09-10 |
| See Less, See Right: Bi-directional Perceptual Shaping For Multimodal Reasoning | ✓ Link | 76.3 | BiPS (Qwen3-VL-8B-Thinking) | 2025-12-26 |
| GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning | ✓ Link | 75.9 | GLM-4.6V | 2025-07-01 |
| Kwai Keye-VL-2.0 Technical Report | ✓ Link | 75.9 | Keye-VL-2.0-30B-A3B | 2026-06-09 |
| Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving | ✓ Link | 75.87 | Qwen-Drive-1.0-SFT | 2026-08-31 |
| ERNIE 5.0 Technical Report | | 75.54 | ERNIE 5.0 | 2026-02-04 |
| Kwai Keye-VL Technical Report | ✓ Link | 75.5 | Keye-VL-8B | 2025-07-02 |
| Qwen3-Omni Technical Report | ✓ Link | 75.5 | Qwen3-Omni-Flash-Thinking | 2025-09-22 |
| iVGR: Internalizing Visually Grounded Reasoning for MLLMs with Reinforcement Learning | ✓ Link | 75.1 | iVGR-Qwen3-VL-32B (ours) | 2026-05-29 |
| EXAONE 4.5 Technical Report | ✓ Link | 74.9 | EXAONE 4.5 (Reasoning) | 2026-04-09 |
| Context-Aware RL for Agentic and Multimodal LLMs | ✓ Link | 74.8 | ContextRL (trained from Qwen3-VL-8B) | 2026-06-15 |
| V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning | ✓ Link | 74.4 | V-Zero-4B (Ours) | 2026-06-24 |
| OpenVLThinkerV2: A Generalist Multimodal Reasoning Model for Multi-domain Visual Tasks | ✓ Link | 73.8 | OpenVLThinkerV2 (Ours) | 2026-04-09 |
| Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoning | | 73.5 | Scaffolding Encoder+RL | 2026-09-29 |
| EvoCUA: Evolving Computer Use Agents via Learning from Scalable Synthetic Experience | ✓ Link | 73.20 | EvoCUA-32B | 2026-01-22 |
| MiniCPM-o 4.5: Towards Real-Time Full-Duplex Omni-Modal Interaction | ✓ Link | 73.1 | MiniCPM-o 4.5-Instruct | 2026-04-30 |
| From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models | ✓ Link | 73.1 | Qwen3-VL-8B (staged training) | 2026-05-19 |
| Near-Future Policy Optimization | | 72.63 | AutoNPO | 2026-04-22 |
| InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models | ✓ Link | 72.5 | InternVL3-78B | 2025-04-14 |
| Ovis2.5 Technical Report | ✓ Link | 72.4 | Ovis2.5-9B | 2025-08-15 |
| MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe | ✓ Link | 72.1 | MiniCPM-V 4.5 | 2025-09-16 |
| TACO: Tool-Augmented Credit Optimization for Agentic Tool Use | | 72.1 | TACO (Qwen3-VL-8B, 8B) | 2026-06-29 |
| Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs | ✓ Link | 71.6 | PVM-8B (SFT+GRPO) | 2026-05-01 |
| Awaking Spatial Intelligence in Unified Multimodal Understanding and Generation | ✓ Link | 71.3 | JoyAI-Image-Und | 2026-05-05 |
| SwimBird: Eliciting Switchable Reasoning Mode in Hybrid Autoregressive MLLMs | ✓ Link | 71.2 | SwimBird | 2026-02-05 |
| Dynamic Important Example Mining for Reinforcement Finetuning | ✓ Link | 71.2 | Qwen2.5-VL-32B + DIEM | 2026-08-29 |
| Qwen2.5-VL Technical Report | ✓ Link | 70.8 | Qwen2.5-VL-72B | 2025-02-19 |
| OneThinker: All-in-one Reasoning Model for Image and Video | ✓ Link | 70.6 | OneThinker-8B | 2025-12-02 |
| Kimi-VL Technical Report | ✓ Link | 70.4 | Kimi-VL-A3B-Thinking-2506 | 2025-04-10 |
| VIVAS: Vitalizing Visual Perception in VLM Pre-training via Vision-language Unified Autoregressive Supervision | | 70.4 | VIVAS | 2026-09-08 |
| Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space | ✓ Link | 70.0 | Qwen3-VL-8B + DMLR | 2025-12-14 |
| AnE: Pushing the Reasoning Frontier of Multimodal LLMs via Anchor Evolution | ✓ Link | 69.9 | AnE-3rd | 2026-05-25 |
| DeepAlign: Mitigating Modality Conflict through Modality-Specific Alignment | | 69.8 | DeepAlign (InternVL3-8B) | 2026-06-03 |
| Infinity-Parser2-Pro | | 69.66 | Infinity-Parser2-Pro | 2026-05-11 |
| LongCat-Next: Lexicalizing Modalities as Discrete Tokens | ✓ Link | 69.3 | LongCat-Next (A3B, 68.5B) | 2026-03-29 |
| Unfold The World: Factorize 4D Properties in Reinforcing Spatial Reasoning | ✓ Link | 69.1 | FactoSR-8B-RL | 2026-09-03 |
| CoLT: Teaching Multi-Modal Models to Think with Chain of Latent Thoughts | ✓ Link | 68.9 | CoLT | 2026-06-30 |
| MetaphorStar: Image Metaphor Understanding and Reasoning with End-to-End Visual Reinforcement Learning | ✓ Link | 68.5 | MetaphorStar-32B | 2026-02-11 |
| Semantic Generative Tuning for Unified Multimodal Models | ✓ Link | 68.33 | SGT-BAGEL (7B+7B) | 2026-05-18 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 68.3 | Qwen2-VL-72B | 2024-09-18 |
| UniPath: Adaptive Coordination of Understanding and Generation for Unified Multimodal Reasoning | | 68.07 | UniPath (BAGEL backbone) | 2026-05-12 |
| Step-GUI Technical Report | ✓ Link | 68.0 | Step-GUI-8B | 2025-12-17 |
| Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model | ✓ Link | 67.32 | Mage-VL-4B | 2026-07-27 |
| From Pixels to Words -- Towards Native One-Vision Models at Scale | ✓ Link | 67.3 | NEO-ov 9B | 2026-05-27 |
| Unlocking Token Rewards via Training-Free Reward Attribution | | 67.2 | Qwen2.5-VL-7B-Instruct + GRPO (P2T token reward) (extra data) | 2026-06-03 |
| Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding? | ✓ Link | 67.20 | Robust-U1 | 2026-06-06 |
| LLaVA-UHD v4: What Makes Efficient Visual Encoding in MLLMs? | ✓ Link | 66.9 | LLaVA-UHD v4 | 2026-05-09 |
| SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces | | 66.9 | SpatialOPSD-9B | 2026-10-08 |
| Toward Cognitive Supersensing in Multimodal Large Language Model | ✓ Link | 66.8 | CogSense-8B | 2026-02-02 |
| PointThinker: Point-Incentivized Parallel Thinking for Multimodal Large Language Model | | 66.8 | PointThinker-7B (extra data) | 2026-06-03 |
| Innovator-VL: A Multimodal Large Language Model for Scientific Discovery | ✓ Link | 66.15 | Innovator-VL (8B-Thinking) | 2026-01-27 |
| LLaVA-OneVision: Easy Visual Task Transfer | ✓ Link | 66.1 | LLaVA-OneVision-72B | 2024-08-06 |
| Language-guided Frequency Modulation for Large Vision-Language Models | | 66.1 | LFM (w/ Qwen2.5-VL) | 2026-06-03 |
| MOSS-VL Technical Report | ✓ Link | 66.0 | MOSS-VL-Instruct-0708 | 2026-08-15 |
| VERDICT: Training-Free Step-Wise Verification of Multimodal Reasoning via Disagreement-Aware Consensus | | 65.88 | VERDICT | 2026-08-11 |
| Puzzle Curriculum GRPO for Vision-Centric Reasoning | ✓ Link | 65.2 | Qwen2.5-VL-7B (Mix) | 2025-12-16 |
| Beyond Layer-Wise Merging: Chain-of-Merging for Vision-Language Models | | 65.0 | CoM (Qwen2.5-VL-7B + DeepSeek-R1-Distill-Qwen-7B) | 2026-06-03 |
| LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence | ✓ Link | 64.8 | LLaVA-OneVision-2-8B | 2026-05-25 |
| Phi-4-reasoning-vision-15B Technical Report | ✓ Link | 64.5 | Phi-4-reasoning-vision-15B | 2026-03-04 |
| Perceptio: Perception Enhanced Vision Language Models via Spatial Token Generation | | 64.2 | Perceptio (8B) | 2026-03-19 |
| From Narrow to Panoramic Vision: Attention-Guided Cold-Start Reshapes Multimodal Reasoning | ✓ Link | 64.1 | AVAR-Thinker | 2026-03-04 |
| LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model | ✓ Link | 64.1 | LLaDA2.0-Uni (16B) | 2026-04-22 |
| LASER: A Corrective Lens for LVLMs via Visual Attention Preservation and Sink Suppression | ✓ Link | 64.1 | LASER | 2026-07-02 |
| Qwen2.5-Omni Technical Report | ✓ Link | 64.0 | Qwen2.5-Omni-7B | 2025-03-26 |
| Qwen2.5-VL Technical Report | ✓ Link | 63.9 | Qwen2.5-VL-7B | 2025-02-19 |
| Improving Vision-language Models with Perception-centric Process Reward Models | ✓ Link | 63.8 | Qwen2.5-VL (7B) with Perceval | 2026-04-27 |
| Unlocking Dense Metric Depth Estimation in VLMs | ✓ Link | 63.8 | DepthVLM-8B | 2026-05-15 |
| OpenSpatial: A Principled Data Engine for Empowering Spatial Intelligence | ✓ Link | 63.7 | OpenSpatial-Qwen3-VL-8B | 2026-04-08 |
| PerceptionDLM: Parallel Region Perception with Multimodal Diffusion Language Models | ✓ Link | 63.7 | PerceptionDLM-Base (8B) | 2026-06-17 |
| Harnessing Textual Refusal Directions for Multimodal Safety | ✓ Link | 63.6 | MARS (Molmo2 8B) | 2026-10-16 |
| Beyond Simply Environment Scaling: Designing Effective Environment Distributions for Multimodal Agent Learning | ✓ Link | 63.5 | AES+HDC (Qwen3-VL-8B-Instruct) | 2026-08-06 |
| LoopVL: Recurrent Visual Intelligence | ✓ Link | 63.47 | LoopVL-1B | 2026-09-29 |
| Reinforcement Learning for Multimodal Diffusion Language Models via Bidimensional Trajectory and Thought Optimization | | 63.2 | LLaDA-V + Bi-VRL (ours) | 2026-09-08 |
| Decouple to Generalize: Context-First Self-Evolving Learning for Data-Scarce Vision-Language Reasoning | ✓ Link | 63.0 | DoGe-7B (Iter3) | 2025-12-07 |
| VisionFoundry: Teaching VLMs Visual Perception with Synthetic Images | ✓ Link | 62.7 | MiMo-VL-7B-SFT + VisionFoundry-10K | 2026-04-10 |
| SpatialEvo: Self-Evolving Spatial Intelligence via Deterministic Geometric Environments | ✓ Link | 62.5 | SpatialEvo (Qwen2.5-VL-7B-Instruct) | 2026-04-15 |
| From Pixels to Words -- Towards Native Vision-Language Primitives at Scale | ✓ Link | 62.4 | NEO-9B | 2025-10-16 |
| GGSS: Geodesic-Gated Spherical Steering for Inference-Time Debiasing of Generative Vision-Language Models | ✓ Link | 62.1 | GGSS (Qwen3-VL-4B, gender steering) | 2026-08-26 |
| Kimi-VL Technical Report | ✓ Link | 61.3 | Kimi-VL-A3B-Instruct | 2025-04-10 |
| LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents | ✓ Link | 61.3 | LLaDA-UI (pre-training) | 2026-09-09 |
| HyperVL: An Efficient and Dynamic Multimodal Large Language Model for Edge Devices | | 61.1 | HyperVL ViTL | 2025-12-16 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 60.7 | Qwen2-VL-7B | 2024-09-18 |
| ROSE: Rotate Your Large Language Model to See | | 60.7 | ROSE-7B (extra data) | 2026-06-03 |
| CLEAR: Unlocking Generative Potential for Degraded Image Understanding in Unified Multimodal Models | ✓ Link | 60.67 | CLEAR-RL (Bagel-7B) | 2026-04-06 |
| Forest Before Trees: Latent Superposition for Efficient Visual Reasoning | ✓ Link | 60.27 | Laser (Qwen2.5-VL-7B-Instruct) | 2026-01-11 |
| LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning | ✓ Link | 60.1 | LLaDA-V | 2025-05-22 |
| Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone | ✓ Link | 59.9 | Dream-VL | 2025-12-27 |
| Information-Regularized Attention for Visual-Centric Reasoning | | 58.8 | InternVL2.5-IRA | 2026-07-01 |
| LFM2 Technical Report | ✓ Link | 57.73 | LFM2-VL-3B | 2025-11-28 |
| UniDDT: Unifying Multimodal Understanding and Generation with Decoupled Diffusion Transformer | | 57.7 | VLM-UniDDT | 2026-06-15 |
| Robust-R1: Degradation-Aware Reasoning for Robust Visual Understanding | ✓ Link | 56.86 | Robust-R1 (SFT and RL) | 2025-12-19 |
| Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System | | 56.51 | Native UMM Synergy: Task-decoupled MoT U+G | 2026-09-01 |
| NGM: A Plug-and-Play Training-Free Memory Module for LLMs | ✓ Link | 56.2 | Qwen3-VL-2B + NGM | 2026-05-16 |
| Building and better understanding vision-language models: insights and future directions | ✓ Link | 55.9 | Idefics3-8B | 2024-08-22 |
| InfiniteVL: Synergizing Linear and Sparse Attention for Highly-Efficient, Unlimited-Input Vision-Language Models | ✓ Link | 55.6 | InfiniteVL-Base | 2025-12-09 |
| UVU: Improving Multimodal Understanding via Vision-Language Unified Autoregressive Paradigm | | 55.0 | UVU 3B | 2026-06-03 |
| LaViT: Aligning Latent Visual Thoughts for Multi-modal Reasoning | ✓ Link | 54.07 | LaViT (3B) | 2026-01-15 |
| PixelUMM: Encoder-Free Unified Image and Video Understanding and Generation | ✓ Link | 53.99 | PixelUMM 8B MoT | 2026-09-29 |
| Differences That Matter: Auditing Models for Capability Gap Discovery and Rectification | | 52.4 | Gemma3 + AuditDM | 2025-12-18 |
| Cheers: Decoupling Patch Details from Semantic Representations Enables Unified Multimodal Comprehension and Generation | ✓ Link | 50.9 | Cheers (1.5B) | 2026-03-13 |
| Introducing LFM2.5: The Next Generation of On-Device AI | | 50.67 | LFM2.5-VL-1.6B | 2026-01-05 |
| NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs | | 50.14 | NeuPAT (Qwen2.5-14B-Instruct) (extra data) | 2026-08-08 |
| Ovis: Structural Embedding Alignment for Multimodal Large Language Model | ✓ Link | 49.5 | Ovis-Llama3-8B | 2024-05-31 |
| AutoNeural: Co-Designing Vision-Language Models for NPU Inference | | 49.4 | AutoNeural | 2025-12-02 |
| SVR-R1: Bootstrapping Multi-modal Reasoning with Self-verification in Reinforcement Learning | | 49.3 | SVR-R1 (7B ThinkLite) | 2026-07-13 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 48.0 | Qwen2-VL-2B | 2024-09-18 |
| Preventing Expert Collapse in MoE-dVLMs via Modality-Wise Norm Alignment | | 46.7 | TowerAlign | 2026-09-08 |
| SmolVLM: Redefining small and efficient multimodal models | ✓ Link | 46.0 | SmolVLM 2.2B | 2025-04-07 |
| Boosting Visual Instruction Tuning with Self-Supervised Guidance | ✓ Link | 43.2 | LLaVA-1.5-Qwen2.5-7B + V-GIFT | 2026-04-14 |
| LFM2.5-VL-450M: Structured Visual Intelligence, Edge to Cloud | | 43.00 | LFM2.5-VL-450M | 2026-04-08 |
| GeMoE: Gating Entropy is All You Need for Uncertainty-aware Adaptive Routing in MoE-based Large Vision-Language Models | ✓ Link | 41.1 | GeMoE (extra data) | 2026-06-24 |
| LinguDistill: Recovering Linguistic Ability in Vision- Language Models via Selective Cross-Modal Distillation | | 34.9 | LinguDistill | 2026-04-01 |
| SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion | ✓ Link | 17.0 | SmolDocling | 2025-03-14 |