SciTaRC: leaderboard
Metric: LLM-judge accuracy (%) with chain-of-thought prompting on the 371 expert-authored SciTaRC questions over tables from recent arXiv papers (LaTeX table sources as input), zero-shot; an answer counts only when a Llama-3.3-70B-Instruct judge gives full credit against the gold answer (partial matches count as wrong); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 76.8 | #91 |
| 2 | Grok 4.1 Fast (Reasoning) | 76.8 | #208 (Grok 4.1 Fast) |
| 3 | Grok 4 | 74.4 | #169 |
| 4 | DeepSeek V3.2 (Thinking) | 73.6 | #198 (DeepSeek V3.2) |
| 5 | Kimi K2 (Thinking) | 73.3 | #236 (Kimi K2) |
| 6 | Claude Opus 4.5 | 72 | #79 |
| 7 | DeepSeek V3.2 (Non-reasoning) | 69.3 | #198 (DeepSeek V3.2) |
| 8 | GPT-OSS-120B | 62.5 | #330 |
| 9 | Gemini 2.5 Flash | 60.6 | #237 |
| 10 | Qwen 3 Next 80B A3B | 49.3 | #306 |
| 11 | Kimi K2 | 49.1 | #236 |
| 12 | GPT-4o | 46.9 | #333 |
| 13 | Qwen 2.5 72B Instruct | 42 | #436 |
| 14 | DeepSeek R1 Distill Qwen 32B | 41.5 | #640 |
| 15 | Llama 3.3 70B Instruct | 34.5 | #520 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=scitarc · How It Works · Data refreshed daily, snapshot 2026-10-11.