BrickBench (Overall, BrickAgent) - VQA: leaderboard

Metric: VQA (fraction); published Overall cohort. Source: www.brickben.ch. Saturation forecast: Around January 2027. 11 models tracked.

Top models

#ModelScore
1GPT-6 Astra0.95
2GPT-6.1 Sol0.95
3Claude Opus 5.50.92
4Claude Opus 50.91
5GPT-5.6 Sol0.86
6Gemini 3.8 Flash0.86
7Qwen 3.8 Flash0.85
8DeepSeek V4.1 Flash0.81
9GPT-5.6 Luna0.79
10Muse Spark 1.30.72
11GLM-5.3 Flash0.59

Interactive version: theaggregate.ai/benchmark?slug=brickbench-overall-brickagent-vqa · How It Works · Data refreshed daily, snapshot 2026-10-10.