CharXiv Reasoning
Multimodal · 2024-06-26
CharXiv Reasoning evaluates reasoning over charts drawn from arXiv papers. Scores report answer accuracy on reasoning questions.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3.1 Pro | 89.0 |
| Claude Fable 5 | 88.9 |
| Fugu Ultra v1.0 | 86.6 |
| Gemini 3.8 Flash | 86.2 |
| Claude Mythos Preview | 86.1 |
| Gemini 3.6 Flash | 85.2 |
| Kimi K3 | 84.8 |
| GPT-5.6 Sol | 84.6 |
| Qwen3.8-Flash-Next | 84.6 |
| Gemini 3.7 Flash | 84.5 |
| Gemini 3.5 Flash | 84.2 |
| GPT-5.5 | 84.1 |
| Gemini 3.1 Pro Preview | 83.3 |
| GPT-5.6 Luna | 82.7 |
| MiniMax M3.1 Flash | 82.5 |
| Opus 4.7 | 82.1 |
| GPT-5.2 | 82.1 |
| Grok 4.5 | 81.6 |
| Gemini 3 Pro Preview | 81.4 |
| MiMo-V2.5 | 81.0 |
| Qwen3.5 397B A17B | 80.8 |
| Opus 4.8 | 80.5 |
| Kimi K2.6 | 80.4 |
| Gemini 3 Flash Preview | 80.3 |
| GPT-5.4 Mini | 80.3 |
| o3 | 78.6 |
| Inkling | 78.1 |
| Kimi K2.5 | 77.5 |
| Inkling-Small | 77.4 |
| Sonnet 5 | 77.0 |
| Inkling-Small Preview | 76.7 |
| Gemini 3.5 Flash-Lite | 74.5 |
| Gemini 3.1 Flash-Lite Preview | 73.2 |
| Gemini 3.1 Flash-Lite | 73.2 |
| Qwen3.5-Omni-Plus | 72.5 |
| Sonnet 4.6 | 72.4 |
| O4 Mini | 72.0 |
| Gemini 2.5 Pro | 69.6 |
| GPT-5.1 | 69.5 |
| Sonnet 4.5 | 68.5 |