CharXiv Reasoning (with tools)
Multimodal · 2024-06-26
This CharXiv subtrack evaluates reasoning over scientific-paper figures with Python tools available. Scores report answer accuracy in percentage points.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 93.5 |
| Claude Mythos Preview | 93.2 |
| Kimi K3 | 91.3 |
| Opus 4.7 | 91.0 |
| Opus 4.8 | 89.9 |
| Gemini 3.6 Flash | 89.4 |
| GPT-5.6 Sol | 89.1 |
| GPT-5.5 | 89.0 |
| Sonnet 5 | 88.3 |
| Kimi K2.6 | 86.7 |
| Sonnet 4.6 | 85.1 |
| Gemini 3.5 Flash | 84.9 |
| Inkling-Small Preview | 83.4 |
| Gemini 3.1 Pro Preview | 83.2 |
| Inkling | 82.0 |
| Inkling-Small | 81.3 |
| Opus 4.6 | 78.9 |
| Kimi K2.5 | 78.7 |
| Gemini 3.5 Flash-Lite | 76.5 |
| Gemini 3.1 Flash-Lite | 75.6 |
| GPT-5 Mini | 75.5 |