TableVista - Complex Structure: leaderboard
Accuracy on the complex tables with multi-level headers and merged cells (Web-style rendering) of TableVista (2026), a multimodal table reasoning benchmark in which models answer questions about rendered table images directly with thinking disabled, scored by exact match backed by a semantic-equivalence check. Percent; higher is better.
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 February 2027. 29 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.4 | Non-reasoning | 81.7 | #135 |
| 2 | Qwen 2.5 VL 72B Instruct | 59.7 | #841 | |
| 3 | GPT-5.4 Mini | Non-reasoning | 59.3 | #406 |
| 4 | Gemma 4 31B (IT) | 57.6 | #350 | |
| 5 | Llama 4 Maverick Instruct FP8 | 55.9 | #1023 | |
| 6 | Qwen 2.5 VL 32B Instruct | 54.5 | #1062 | |
| 7 | Qwen 3.5 27B | Non-reasoning | 53 | #375 |
| 8 | Qwen 3.5 122B A10B | Non-reasoning | 52.2 | #346 |
| 9 | Gemma 4 26B A4B (IT) | 50.1 | #556 | |
| 10 | Llama 4 Scout Instruct | 49.6 | #1579 | |
| 11 | InternVL3.5-14B | 48.9 | #1504 | |
| 12 | InternVL3.5-30B-A3B | 47.7 | #1267 | |
| 13 | Qwen 2.5 VL 7B Instruct | 47.5 | #1902 | |
| 14 | Qwen 3 VL 30B A3B Instruct | 46.8 | #895 | |
| 15 | Molmo2-8B | 45.2 | #1621 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=tablevista-complex-structure · How It Works · Data refreshed daily, snapshot 2026-11-21.