BabyVision
Can the model extract and reason over information in images or documents?
Top rankings
One row per model, using its best reported score across effort settings.
| Rank | Model | Best score | Best reported setting |
|---|---|---|---|
| 1 | DeepSeek V4.1 FlashDeepSeek | 89.6% | Reported configurationDeepSeek report1 value · 1 reportSep 10, 2026 · source |
| 2 | GPT-5.5OpenAI | 83.6% | xhigh effortMeta report1 value · 1 reportJul 9, 2026 · source |
| 3 | Claude Opus 4.8Anthropic | 81.2% | max effortMeta report1 value · 1 reportJul 9, 2026 · source |
| 4 | Muse Spark 1.1Meta | 76.3% | xhigh effortMeta report1 value · 1 reportJul 9, 2026 · source |
| 5 | Gemini 3.1 ProGoogle | 51.5% | high effortMeta report1 value · 1 reportJul 9, 2026 · source |
| 6 | Muse Spark 1.0Meta | 39.9% | Reported configurationMeta report1 value · 1 reportJul 9, 2026 · source |
Effort curve
Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.
Definition and comparison boundarypublic methodology
Can the model extract and reason over information in images or documents?
Visual perception and multimodal reasoning under the reported protocol. Higher is better. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
DeepSeek's official V4.1-Flash release table; exact source label BabyVision (w/tools) retained. 388 questions; gpt-oss-120B high-reasoning judge; with code execution.
Method / source