PCB-Bench - Task 3 F1: leaderboard
Harmonic mean of precision and recall over the design facts a model states in PCB-Bench task 3.
Metric: F1 score (%). Source: digailab.github.io. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3-78B | 88.49 | #352 |
| 2 | Qwen 3 VL 8B Instruct | 86.75 | #420 |
| 3 | GPT-5 | 85.91 | #98 |
| 4 | Gemini 2.5 Flash Lite | 85.63 | #422 |
| 5 | Llama 4 Maverick | 85.11 | #464 |
| 6 | GPT-5 Nano | 85.07 | #425 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 84.05 | #238 (Qwen 3 VL 235B A22B) |
| 8 | Gemini 2.5 Pro | 80.89 | #152 |
Interactive version: theaggregate.ai/benchmark?slug=pcb-bench-task-3-f1 · How It Works · Data refreshed daily, snapshot 2026-10-20.