CharXiv Reasoning (with tools)Browse 284 families
CharXiv Reasoning (with tools)
Can the model extract and reason over information in images or documents?
Version / measure
- Domain
- Vision
- Ranked models
- 9
- Reported values
- 12
- Reports
- 3 · lab
- Direction
- Higher is better
- Status
- Version not specified
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | Claude Opus 4.8Anthropic | maxcost n/a1 value · Jul 9 · source | 89.9% |
| 2 | Gemini 3.6 FlashGoogle | reportedcost n/a2 values · Aug 13 · source | 89.4% |
| 3 | Muse Spark 1.0Meta | reportedcost n/a1 value · Jul 9 · source | 88.9% |
| 4 | Gemini 3.7 FlashGoogle | reportedcost n/a1 value · Aug 13 · source | 88.7% |
| 5 | Muse Spark 1.1Meta | xhighcost n/a1 value · Jul 9 · source | 88.4% |
| 6 | Claude Sonnet 5Anthropic | reportedcost n/a2 values · Aug 13 · source | 88.3% |
| 7 | Gemini 3.5 FlashGoogle | reportedcost n/a1 value · Jul 21 · source | 84.9% |
| 8 | GPT-5.5OpenAI | xhighcost n/a1 value · Jul 9 · source | 84.8% |
| 9 | Gemini 3.1 ProGoogle | reportedcost n/a2 values · Jul 21 · source | 83.2% |
Effort curve
Score vs. reported cost. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
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.
With tools. Gemini's tool-enabled setup uses search and code execution. 1,000 questions; gpt-oss-120B high-reasoning judge; with code execution.
Method / source ↗