TableVista - Missing: leaderboard
Metric: Accuracy (%) on the 3,000 TableVista questions under the Missing rendering (arbitrary cells masked while the answer stays derivable), 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) | 84.8 |
| 2 | Gemma 4 31B (IT) | 75.7 |
| 3 | Qwen 2.5 VL 72B Instruct | 71.3 |
| 4 | Qwen 2.5 VL 32B Instruct | 67.3 |
| 5 | Llama 4 Maverick Instruct FP8 | 66.9 |
| 6 | GPT-5.4 Mini (Non-reasoning) | 66.9 |
| 7 | Gemma 4 26B A4B (IT) | 65.6 |
| 8 | Llama 4 Scout Instruct | 64.9 |
| 9 | Qwen 3.5 27B (Non-reasoning) | 63.8 |
| 10 | Qwen 3.5 122B A10B (Non-reasoning) | 63.3 |
| 11 | Qwen 2.5 VL 7B Instruct | 61.8 |
| 12 | Gemma 3 27B (IT) | 61.1 |
| 13 | Qwen 3 VL 30B A3B Instruct | 60.2 |
| 14 | Qwen 3 VL 8B Instruct | 57.7 |
| 15 | Molmo2-8B | 57.5 |
Interactive version: theaggregate.ai/benchmark?slug=tablevista-missing · How It Works · Data refreshed daily, snapshot 2026-10-07.