Multimodal Understanding on VMCBench
87.7Overall ScoreTimage
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| TimageModel Scale Category=Efficient Models (≤8B), Base Model=Qwen2.5-VL-7B, Ensemble voting (K)=52026.06 | 87.7 | — | — | — | — | — | 91.3 | — | — | — | — | 75.8 | 98.5 | 92.6 | |
| Qwen3-8BModel Scale Category=Efficient Models (≤8B)2026.06 | 87.3 | — | — | — | — | — | 87.3 | — | — | — | — | 71.8 | 96.4 | 89.5 | |
| Qwen2.5-VL-14BModel Scale Category=Open-Source Mid-Scale Models (10B–40B)2026.06 | 85.9 | — | — | — | — | — | 89.8 | — | — | — | — | 72.6 | 96.5 | 89.9 | |
| Qwen2-VL-72BModel Scale Category=Proprietary / Large-Scale Models (>30B)2026.06 | 85 | — | — | — | — | — | 88.5 | — | — | — | — | 72.6 | 96.8 | 90.1 | |
| RC-DPOEvaluation Protocol=MM-Eureka2026.05 | 81.55 | 59.62 | 74.26 | 80.41 | 91.44 | 80.25 | 84.73 | 97.53 | 65.08 | 98.44 | 90.12 | — | — | — | |
| OPA-DPOEvaluation Protocol=MM-Eureka2026.05 | 81.39 | 58.89 | 73.76 | 77.9 | 90.51 | 80.74 | 84.09 | 96.85 | 64.61 | 98.44 | 89.65 | — | — | — | |
| VCDEvaluation Protocol=MM-Eureka2026.05 | 80.95 | 57.93 | 74.63 | 79.73 | 89.81 | 78.52 | 84.25 | 96.4 | 65.08 | 90.87 | 89.18 | — | — | — | |
| VanillaEvaluation Protocol=MM-Eureka2026.05 | 80.88 | 57.93 | 73.27 | 79.04 | 90.05 | 76.05 | 83.52 | 96.63 | 63.66 | 91.76 | 88.47 | — | — | — | |
| RLAIF-VEvaluation Protocol=MM-Eureka2026.05 | 80.85 | 57.45 | 69.31 | 80.87 | 90.51 | 79.26 | 84.21 | 96.4 | 65.8 | 95.32 | 88.71 | — | — | — | |
| SIDEvaluation Protocol=MM-Eureka2026.05 | 80.72 | 58.89 | 74.22 | 78.59 | 89.58 | 79.26 | 83.99 | 95.51 | 61.76 | 91.09 | 88.71 | — | — | — | |
| GPT-4oModel Scale Category=Proprietary / Large-Scale Models (>30B)2026.06 | 80.3 | — | — | — | — | — | 85.2 | — | — | — | — | 66.9 | 96.4 | 83.1 | |
| Qwen2.5-VL-7B(Base)Model Scale Category=Efficient Models (≤8B)2026.06 | 79.1 | — | — | — | — | — | 85.6 | — | — | — | — | 68.2 | 95.3 | 86.5 | |
| Molmo-72BModel Scale Category=Proprietary / Large-Scale Models (>30B)2026.06 | 78.7 | — | — | — | — | — | 82.9 | — | — | — | — | 66.6 | 94.7 | 81.1 | |
| Qwen2-VL-7BModel Scale Category=Efficient Models (≤8B)2026.06 | 78.1 | — | — | — | — | — | 84.5 | — | — | — | — | 62.7 | 96.4 | 80.1 | |
| Claude-3.5-SonnetModel Scale Category=Proprietary / Large-Scale Models (>30B)2026.06 | 77.8 | — | — | — | — | — | 81.3 | — | — | — | — | 62.8 | 93.4 | 84.6 | |
| Cambrian-34BModel Scale Category=Open-Source Mid-Scale Models (10B–40B)2026.06 | 77 | — | — | — | — | — | 83.7 | — | — | — | — | 65.9 | 95.7 | 73.3 | |
| VILA1.5-40BModel Scale Category=Open-Source Mid-Scale Models (10B–40B)2026.06 | 74.7 | — | — | — | — | — | 82.5 | — | — | — | — | 65.3 | 93.2 | 67.4 | |
| CogVLM2-19BModel Scale Category=Open-Source Mid-Scale Models (10B–40B)2026.06 | 71.4 | — | — | — | — | — | 78.1 | — | — | — | — | 55.6 | 92.3 | 72.6 | |
| Phi-3-VisionModel Scale Category=Efficient Models (≤8B)2026.06 | 70.3 | — | — | — | — | — | 74.1 | — | — | — | — | 56.4 | 90.6 | 73.8 | |
| Cambrian-8BModel Scale Category=Efficient Models (≤8B)2026.06 | 69.6 | — | — | — | — | — | 77.9 | — | — | — | — | 56.4 | 91 | 65.4 | |
| Molmo-7B-DModel Scale Category=Efficient Models (≤8B)2026.06 | 69.5 | — | — | — | — | — | 73.2 | — | — | — | — | 55.5 | 91.7 | 72.1 | |
| LLaVA1.5-7BModel Scale Category=Efficient Models (≤8B)2026.06 | 51.8 | — | — | — | — | — | 63.6 | — | — | — | — | 44.7 | 74 | 35 |