Visual Search on HR-Bench 4K
81.25AccuracyDynamo + RL
Evaluation Results
| Method | Links | |
|---|---|---|
| Dynamo + RLBackbone=Qwen2.5-VL-7B2026.06 | 81.25 | |
| Qwen2.5-VL-72BModel Category=Open-Source MLLMs2025.12 | 79.4 | |
| InterSketch2026.05 | 78.5 | |
| PyVision-ImageTooling Type=Dynamic Tooling2026.02 | 78.1 | |
| DeepEyes-v2Tooling Type=Dynamic Tooling2026.02 | 77.9 | |
| DeepEyeV2-7B2026.05 | 77.9 | |
| Mini-o3Tooling Type=Static Toolset2026.02 | 77.5 | |
| Thyme-VL-7BModel Category=Open-Source MLLMs with Tools2025.12 | 77 | |
| ThymeTooling Type=Dynamic Tooling2026.02 | 77 | |
| Thyme-7B2026.05 | 77 | |
| RLBackbone=Qwen2.5-VL-7B2026.06 | 76.88 | |
| CodeVTooling Type=Dynamic Tooling2026.02 | 76.1 | |
| CodeVisionTooling Type=Dynamic Tooling2026.02 | 75.6 | |
| InternVL3-78BModel Category=Open-Source MLLMs2025.12 | 75.5 | |
| CodeDance-7BModel Category=Open-Source MLLMs with Tools2025.12 | 75.2 | |
| CodeDanceTooling Type=Dynamic Tooling2026.02 | 75.2 | |
| DynamoBackbone=Qwen2.5-VL-7B2026.06 | 75.12 | |
| DeepEyesTooling Type=Static Toolset2026.02 | 75.1 | |
| Qwen2.5-VL-32BModel Category=Open-Source MLLMs2025.12 | 74.8 | |
| Deepeyes-7BModel Category=Open-Source MLLMs with Tools, reproduced=true2025.12 | 74.8 | |
| Pixel-ReasonerTooling Type=Static Toolset2026.02 | 74 | |
| Pixel Reasoner-7BModel Category=Open-Source MLLMs with Tools2025.12 | 72.9 | |
| InternVL2.5-8BModel Category=Open-Source MLLMs2025.12 | 72 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.02 | 71.6 | |
| Dynamo + RLBackbone=Qwen2.5-VL-3B2026.06 | 71.25 | |
| InternVL3-8BModel Category=Open-Source MLLMs2025.12 | 70.5 | |
| DynamoBackbone=Qwen2.5-VL-3B2026.06 | 70.5 | |
| DirectBackbone=Qwen2.5-VL-7B2026.06 | 70.5 | |
| Qwen3-VL-8B2026.05 | 70 | |
| Qwen2.5-VL-7BModel Category=Open-Source MLLMs2025.12 | 69 | |
| Qwen2.5-VL-7B2026.05 | 68.8 | |
| Llava-OneVision-7BModel Category=Open-Source MLLMs2025.12 | 68.5 | |
| RLBackbone=Qwen2.5-VL-3B2026.06 | 66.5 | |
| Llava-OneVision-72BModel Category=Open-Source MLLMs2025.12 | 66.3 | |
| GPT-4oModel Category=Closed-Source MLLMs2025.12 | 65 | |
| DirectBackbone=Qwen2.5-VL-3B2026.06 | 63.62 | |
| ThinkMorph-7B2026.05 | 54 |