LLM Stats (TextVQA): leaderboard
28,408 questions requiring reading and reasoning about text visible in natural images like signs, labels, and storefronts.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 84.9 | #654 |
| 2 | Qwen2.5-Omni-7B | 84.4 | #624 |
| 3 | Nova Pro | 81.5 | #569 |
| 4 | Nova Lite | 80.2 | #754 |
| 5 | Phi-4 Multimodal Instruct | 75.6 | #906 |
| 6 | Llama 3.2 90B Instruct | 73.5 | #601 |
| 7 | Gemma 3 12B | 67.7 | #671 |
| 8 | Gemma 3 27B | 65.1 | #607 |
| 9 | Gemma 3 4B | 57.8 | #1089 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-textvqa · How It Works · Data refreshed daily, snapshot 2026-10-20.