CADEngBench - Assembly Grounding (Typed@1): leaderboard
The assembly-grounding task of CADEngBench (MBZUAI, 2026): given two B-Rep bodies, the model ranks joint hypotheses, each a joint family and one entity from each body. Typed@1 is the percentage of 120 tasks whose top prediction matches the recorded entity pair, joint family and body roles; higher is better.
Metric: Typed top-1 accuracy (%; share of 120 assembly relation tasks between two source B-Rep bodies; one response per task whose rank-1 prediction selects the recorded B-Rep entity pair with the correct joint family and body roles). Source: arxiv.org. Saturation forecast: Around October 2028. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) (Minimal) | 23.3 | #144 (Gemini 3 Flash (Preview)) |
| 2 | Kimi K2.5 (Non-reasoning) | 15 | #256 (Kimi K2.5) |
| 3 | Claude Sonnet 4.5 | 14.2 | #250 |
| 4 | Llama 4 Maverick | 14.2 | #1099 |
| 5 | Qwen 3.5 35B A3B (Non-reasoning) | 14.2 | #482 (Qwen 3.5 35B A3B) |
| 6 | GPT-5.2 (2025-12-11) (Non-reasoning) | 12.5 | |
| 7 | GLM-4.6V (Non-reasoning) | 10 | #701 (GLM-4.6V) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cadengbench-assembly-grounding-typed-1 · How It Works · Data refreshed daily, snapshot 2026-11-13.