PEC-Home (RAG) - Multi-User Level 3: leaderboard
Metric: Execution accuracy (%; 1,068 smart-home commands at ellipsis level 3 of 4 (progressively elliptical), functionally correct device operations; dialogue history retrieved from a vector store, top-3). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V3 | 76.78 |
| 2 | GPT-4o | 76.05 |
| 3 | Qwen 2.5 14B Instruct | 69.19 |
| 4 | Qwen 2.5 72B Instruct | 69.1 |
| 5 | Qwen 2.5 32B Instruct | 68.07 |
| 6 | Qwen 2.5 7B Instruct | 64.04 |
| 7 | Mistral 7B Instruct (v0.3) | 63.76 |
| 8 | Gemma 2 27B (IT) | 62.27 |
| 9 | Llama 3.1 8B Instruct | 58.71 |
| 10 | Gemma 2 9B (IT) | 55.24 |
Interactive version: theaggregate.ai/benchmark?slug=pec-home-rag-multi-user-level-3 · How It Works · Data refreshed daily, snapshot 2026-09-26.