VEHBench - Specification Triage: leaderboard
VEHBench (2026) is a stage-local diagnostic benchmark for LLM-assisted design of vibration energy harvesters for battery-free IoT, scored against a physics oracle. Composite score from 0 to 100 for deciding whether to propose a design, flag an infeasible brief or ask for missing information; higher is better.
Metric: P1 composite (0-100): weighted combination of macro-F1, action-conditioned, missing-information and infeasibility detection scores and subtype F1 for deciding whether to propose a design, abstain as infeasible or request missing information on a harvester brief, x100; LLM-assisted vibration energy harvester design against a physics oracle; one complete run per model. Source: arxiv.org. Saturation forecast: Not forecast. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Max | 57.4 | #403 |
| 2 | Gemini 3.1 Pro (Preview) | 54.9 | #94 |
| 3 | O4 Mini | 51.8 | #353 |
| 4 | DeepSeek R1 | 50.4 | #473 |
| 5 | GPT-5.4 | 42.8 | #123 |
| 6 | Hy3-preview (Reasoning) | 42.5 | #316 (Hy3-preview) |
| 7 | DeepSeek V3 | 36.9 | #716 |
| 8 | Llama 3.3 70B | 36.1 | #1498 |
| 9 | MiMo-V2.5-Pro (Non-reasoning) | 31.7 | #279 (MiMo-V2.5-Pro) |
| 10 | Qwen 3.6 Plus | 30.6 | #222 |
| 11 | DeepSeek V4 Pro (Non-reasoning) | 30.6 | #188 (DeepSeek V4 Pro) |
| 12 | Claude Sonnet 4.6 | 20.7 | #161 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=vehbench-specification-triage · How It Works · Data refreshed daily, snapshot 2026-11-13.