PCB-Bench - Task 3 SBERT: leaderboard
PCB-Bench task-3 free-form design-explanation generation scored by SBERT sentence-embedding similarity.
Metric: SBERT similarity (%). Source: digailab.github.io. Saturation forecast: Around July 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B (Thinking) | 61.68 | #238 (Qwen 3 VL 235B A22B) |
| 2 | GPT-5 | 61.34 | #98 |
| 3 | Qwen 3 VL 8B Instruct | 57.6 | #420 |
| 4 | Gemini 2.5 Pro | 56.26 | #152 |
| 5 | GPT-5 Nano | 53.58 | #425 |
| 6 | Llama 4 Maverick | 51.01 | #464 |
| 7 | Gemini 2.5 Flash Lite | 50.83 | #422 |
| 8 | InternVL3-78B | 48.5 | #352 |
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-sbert · How It Works · Data refreshed daily, snapshot 2026-10-20.