TableVista - Complex Structure: leaderboard
Metric: Accuracy (%) on the 1,000 complex tables (multi-level headers, merged cells) of the 3,000 TableVista questions, Web rendering, direct-output prompt with thinking disabled (reasoning effort none for the GPT models), exact match with a GPT-5-mini semantic-equivalence check on exact-match failures; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (Non-reasoning) | 81.7 |
| 2 | Qwen 2.5 VL 72B Instruct | 59.7 |
| 3 | GPT-5.4 Mini (Non-reasoning) | 59.3 |
| 4 | Gemma 4 31B (IT) | 57.6 |
| 5 | Llama 4 Maverick Instruct FP8 | 55.9 |
| 6 | Qwen 2.5 VL 32B Instruct | 54.5 |
| 7 | Qwen 3.5 27B (Non-reasoning) | 53 |
| 8 | Qwen 3.5 122B A10B (Non-reasoning) | 52.2 |
| 9 | Gemma 4 26B A4B (IT) | 50.1 |
| 10 | Llama 4 Scout Instruct | 49.6 |
| 11 | Qwen 2.5 VL 7B Instruct | 47.5 |
| 12 | Qwen 3 VL 30B A3B Instruct | 46.8 |
| 13 | Molmo2-8B | 45.2 |
| 14 | Gemma 3 27B (IT) | 44.6 |
| 15 | Qwen 3 VL 8B Instruct | 44.1 |
Interactive version: theaggregate.ai/benchmark?slug=tablevista-complex-structure · How It Works · Data refreshed daily, snapshot 2026-10-07.