PCB-Bench - Task 3 Precision: leaderboard
Precision of the design facts a model states in PCB-Bench task 3 — how much of what it asserts about the board is correct.
Metric: Precision (%). Source: digailab.github.io. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 4 Maverick | 91.23 | #464 |
| 2 | InternVL3-78B | 90.44 | #352 |
| 3 | Gemini 2.5 Pro | 89.44 | #152 |
| 4 | GPT-5 Nano | 87.39 | #425 |
| 5 | Gemini 2.5 Flash Lite | 86.87 | #422 |
| 6 | GPT-5 | 84.93 | #98 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 84.93 | #238 (Qwen 3 VL 235B A22B) |
| 8 | Qwen 3 VL 8B Instruct | 84.63 | #420 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pcb-bench-task-3-precision · How It Works · Data refreshed daily, snapshot 2026-10-20.