DrawingVQA: leaderboard
DrawingVQA (2026) asks multimodal LLMs questions about real Issued for Construction drawings, which mix abstract geometry, symbols, schedules, annotations and engineering text. Accuracy in percent over all questions, higher is better.
Metric: Accuracy (%; all 92 expert-curated questions on 33 Issued for Construction drawings, multiple choice and open-ended). Source: arxiv.org. Saturation forecast: Around February 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 77.2 | #113 |
| 2 | Gemini 2.5 Pro | 71.7 | #266 |
| 3 | Gemini 2.5 Flash | 66.3 | #453 |
| 4 | Qwen 3 VL 32B Instruct | 58.7 | #585 |
| 5 | O3 (2025-04-16) | 58.7 | #216 |
| 6 | Claude Sonnet 4.5 | 57.6 | #250 |
| 7 | Claude Haiku 4.5 | 54.3 | #568 |
| 8 | Qwen 3 VL 8B Instruct | 53.3 | #981 |
| 9 | Llama 4 Scout Instruct | 53.3 | #1543 |
| 10 | GPT-4o (2024-08-06) | 48.9 | #740 |
| 11 | Qwen 3 VL 30B A3B Instruct | 47.8 | #871 |
| 12 | Phi-4 Multimodal Instruct | 40.2 | #3882 |
| 13 | Llama 3.2 11B Instruct | 39.1 | #4991 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=drawingvqa · How It Works · Data refreshed daily, snapshot 2026-11-13.