Counting on CountBench
95.1AccuracyMolmoPoint-8B-O-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MolmoPoint-8B-O-7BAccess=Fully open2026.03 | 95.1 | — | |
| Molmo2-4BAccess=Fully open2026.03 | 93.9 | — | |
| SAM3Encoder=Perception Encoder2026.01 | 93.8 | — | |
| SAM 32025.11 | 93.8 | 0.12 | |
| Molmo2-8BAccess=Fully open2026.03 | 93.7 | — | |
| Qwen2.5-VLParameters=72B2025.02 | 93.6 | — | |
| Qwen2.5-VL-72BModel Category=Open-Source MLLMs2025.12 | 93.6 | — | |
| Qwen2.5-VL-72B2026.04 | 93.6 | — | |
| BAGELModel Category=Unified Models2026.04 | 93.2 | — | |
| MolmoPoint-8BAccess=MolmoPoint2026.03 | 92.9 | — | |
| Qwen3-VL (Thinking)Thinking Mode=true, Parameters=235B-A22B2026.01 | 92.46 | — | |
| Molmo-72B2025.11 | 92.4 | 0.27 | |
| Gemini 2.5 Pro2025.11 | 92.4 | 0.24 | |
| GLM-4.6VParameters=106B-A12B2026.01 | 92.06 | — | |
| Seed-1.5-VL (Thinking)Thinking Mode=true2026.01 | 91.85 | — | |
| PLM-8BAccess=Fully open2026.03 | 91.8 | — | |
| LLaDA-oModel Category=Unified Models2026.04 | 91.7 | — | |
| MolmoParameters=72B2025.02 | 91.2 | — | |
| CodeDance-7BModel Category=Open-Source MLLMs with Tools2025.12 | 91.2 | — | |
| GPT-5Access=API call only2026.03 | 90.8 | — | |
| Gemini 2.5 ProAccess=API call only2026.03 | 90.8 | — | |
| Qwen3-VL-8BAccess=Open weights only2026.03 | 90.4 | — | |
| GLM-4.6V FlashNumber of Parameters=9B2026.01 | 90.22 | — | |
| Claude-4-SonnetSize=-2026.02 | 90.2 | — | |
| Eagle2.5-8BAccess=Open weights only2026.03 | 90.2 | — | |
| Claude 3.7 SonnetParams=–2025.11 | 90.2 | — | |
| Qwen2.5-VL-7B +PCBackbone Model=Qwen2.5-VL-7B, Training Paradigm (+PC)=Yes2026.03 | 89.9 | — | |
| Claude-3.5 Sonnet2025.02 | 89.7 | — | |
| Gemini2.5-ProSize=-2026.02 | 89.7 | — | |
| Claude-3.5 Sonnet2026.04 | 89.7 | — | |
| Qwen2.5-VL-7B AIFmethod=Adaptive Information Flow2026.04 | 89.5 | — | |
| Qwen2.5-VL-7BBackbone Model=Qwen2.5-VL-7B, Training Paradigm (+PC)=No2026.03 | 89.3 | — | |
| MiMo-VL-7B +PCBackbone Model=MiMo-VL-7B, Training Paradigm (+PC)=Yes2026.03 | 89.3 | — | |
| Qwen3-VL ThinkingNumber of Parameters=8B2026.01 | 88.85 | — | |
| STEP3-VL-10B (PaCoRe)Reasoning Strategy=PaCoRe, Parameters=10B2026.01 | 88.8 | — | |
| STEP3-VL-10BNumber of Parameters=10B2026.01 | 88.75 | — | |
| STEP3-VL-10B (SeRe)Reasoning Strategy=SeRe, Parameters=10B2026.01 | 88.75 | — | |
| Youtu-VLModel Size=4B2026.01 | 88.6 | — | |
| GLM-4.1V-9BAccess=Open weights only2026.03 | 88 | — | |
| GPT-4o2025.02 | 87.9 | — | |
| GPT-4oModel Category=Closed-Source MLLMs2025.12 | 87.9 | — | |
| GPT-4o2026.04 | 87.9 | — | |
| Qwen2.5-VL-32BModel Category=Open-Source MLLMs2025.12 | 87.8 | — | |
| Gemini-2.5 (Pro)Model Tier=Pro2026.01 | 87.78 | — | |
| MiMo-EmbodiedParams=7B2025.11 | 87.37 | — | |
| Claude Sonnet 4.5Access=API call only2026.03 | 87.3 | — | |
| GPT-5 miniAccess=API call only2026.03 | 87.1 | — | |
| PLM-3BAccess=Fully open2026.03 | 87.1 | — | |
| Qwen2.5-VL-7B2026.04 | 87.1 | — | |
| MiMo-VL-7BBackbone Model=MiMo-VL-7B, Training Paradigm (+PC)=No2026.03 | 87 | — | |
| Qwen2.5-VLSize=72B2026.02 | 86.7 | — | |
| Qwen2-VL-72B2025.11 | 86.7 | 0.28 | |
| Gemini 2.5 FlashAccess=API call only2026.03 | 86.7 | — | |
| LLaDA2.0-UniModel Category=Unified Models2026.04 | 86 | — | |
| GPT-4oSize=-2026.02 | 85.7 | — | |
| GPT-4oParams=–2025.11 | 85.7 | — | |
| Gemini 1.5-Pro2025.02 | 85.5 | — | |
| Qwen3-VL-4BAccess=Open weights only2026.03 | 85.5 | — | |
| Visual Para-ThinkerSize=7B2026.02 | 85.4 | — | |
| MiMo-VLParams=7B2025.11 | 85.34 | — | |
| Qwen2.5-VL-7BModel Category=Specialist VLMs2026.04 | 84.9 | — | |
| Thyme-VL-7BModel Category=Open-Source MLLMs with Tools, reproduced=true2025.12 | 84.8 | — | |
| GPT5-miniSize=-2026.02 | 84.4 | — | |
| MiniCPM-V-4.5-8BAccess=Open weights only2026.03 | 83.9 | — | |
| SequentialSize=7B2026.02 | 83.7 | — | |
| MiMo-VL RL-2508Number of Parameters=7B2026.01 | 83.6 | — | |
| Majority voting@4Size=7B2026.02 | 83.2 | — | |
| Qwen2.5-VLSize=7B2026.02 | 83.1 | — | |
| DINO-X2025.11 | 82.9 | 0.62 | |
| Llava-OneVision-7BModel Category=Open-Source MLLMs2025.12 | 82.3 | — | |
| InternVL3.5-4BAccess=Open weights only2026.03 | 82.2 | — | |
| InternVL 3.5Number of Parameters=8B2026.01 | 82.18 | — | |
| Keye-VL-1.5-8BAccess=Open weights only2026.03 | 81.6 | — | |
| Deepeyes-7BModel Category=Open-Source MLLMs with Tools, reproduced=true2025.12 | 80.4 | — | |
| InternVL3-8BModel Category=Open-Source MLLMs2025.12 | 80.3 | — | |
| InternVL3Params=8B2025.11 | 80 | — | |
| InternVL3.5-8BAccess=Open weights only2026.03 | 79.6 | — | |
| InternVL-3.5Model Size=4B2026.01 | 79.4 | — | |
| Qwen2.5-VL-3B +PCBackbone Model=Qwen2.5-VL-3B, Training Paradigm (+PC)=Yes2026.03 | 78.8 | — | |
| Qwen3-VLModel Size=4B2026.01 | 78.4 | — | |
| Qwen2.5-VL-3BBackbone Model=Qwen2.5-VL-3B, Training Paradigm (+PC)=No2026.03 | 77.8 | — | |
| Qwen2.5-VL-7BModel Category=Open-Source MLLMs2025.12 | 76.5 | — | |
| MiniCPM-V 4.5Prompting Strategy=Task-Aware2026.01 | 76.4 | — | |
| LLaDA-VModel Category=Specialist VLMs2026.04 | 75.1 | — | |
| Qwen2.5-VLParams=7B2025.11 | 74.1 | — | |
| Visual Para-ThinkerSize=3B2026.02 | 72.6 | — | |
| InternVL2.5Parameters=78B2025.02 | 72.1 | — | |
| SequentialSize=3B2026.02 | 70.4 | — | |
| Majority voting@4Size=3B2026.02 | 69.8 | — | |
| Qwen2.5-VLSize=3B2026.02 | 69.2 | — | |
| InternVL3.5 8BPrompting Strategy=Task-Aware2026.01 | 68.6 | — | |
| InternVL-UModel Category=Unified Models2026.04 | 62.2 | — | |
| InternVL2.5-8BModel Category=Open-Source MLLMs2025.12 | 55.9 | — | |
| MiniCPM-V 4.5Prompting Strategy=General2026.01 | 50.1 | — | |
| LLaVA-1.5-7B AIFmethod=Adaptive Information Flow2026.04 | 50.1 | — | |
| Lumina-DiMOOModel Category=Unified Models2026.04 | 48.4 | — | |
| LLaVA-1.5-7B2026.04 | 47 | — | |
| InternVL3.5 8BPrompting Strategy=General2026.01 | 43.9 | — | |
| Singh et al.2026.01 | 34.2 | — | |
| Zhang et al.2026.01 | 30.7 | — |