BrickBench (Overall, BrickAgent) - Validity: leaderboard
Metric: Validity (fraction); published Overall cohort. Source: www.brickben.ch. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol | 1 |
| 2 | Claude Opus 5 | 1 |
| 3 | GPT-5.6 Luna | 1 |
| 4 | GPT-6 Astra | 1 |
| 5 | GPT-6.1 Sol | 1 |
| 6 | Claude Opus 5.5 | 0.99 |
| 7 | Gemini 3.8 Flash | 0.98 |
| 8 | DeepSeek V4.1 Flash | 0.94 |
| 9 | GLM-5.3 Flash | 0.92 |
| 10 | Qwen 3.8 Flash | 0.9 |
| 11 | Muse Spark 1.3 | 0.8 |
Interactive version: theaggregate.ai/benchmark?slug=brickbench-overall-brickagent-validity · How It Works · Data refreshed daily, snapshot 2026-10-10.