OmniSch - Net Detection: leaderboard
Metric: Net detection F1 (x100) of predicted pin-to-pin connections against the ground-truth nets on the industrial-EDA (Real) schematics of the OmniSch held-out test split, the model converting each PCB schematic image into a spatially weighted netlist graph with one zero-shot prompt, predicted and reference nodes aligned by maximum-weight bipartite matching; higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 122B A10B | 42.37 |
| 2 | Gemini 3.1 Pro (Preview) | 27.06 |
| 3 | GPT-5.2 | 19.87 |
| 4 | Claude Sonnet 4.6 | 17.48 |
| 5 | Ministral 3 14B | 11.51 |
| 6 | GPT-5 Mini | 10.16 |
| 7 | Llama 4 Maverick | 6.72 |
| 8 | Qwen 3 VL 8B Instruct | 0.43 |
Interactive version: theaggregate.ai/benchmark?slug=omnisch-net-detection · How It Works · Data refreshed daily, snapshot 2026-10-07.