PersonalHomeBench - Multiple Choice (Agentic): leaderboard
PersonalHomeBench (2026) tests models as assistants in generated, persona-rich smart homes. Multiple-choice questions probe the same household knowledge with enumerated answers; accuracy in percent, acting through the benchmark household toolbox to gather information. Higher is better.
Metric: Accuracy (%) on 1,000 multiple-choice household questions (exact match on the selected option), over the text-based split of PersonalHomeBench (1,100 generated, manually reviewed households with persona-based occupants and appliances), temperature 0, in the agentic setting (the model must gather household information and act through the benchmark's PersonalHomeTools toolbox, up to 15 turns); each model at its stated reasoning mode; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-OSS-20B | Low | 79.6 | #1183 |
| 2 | GPT-4 | 79.2 | #970 | |
| 3 | Qwen 3 30B | Thinking | 77.6 | |
| 4 | GPT-4o | 75 | #751 | |
| 5 | nemotron-3-nano-30B-a3B | 73.8 | #1946 | |
| 6 | Qwen 3 4B | Non-reasoning | 65.1 | #2348 |
| 7 | Qwen 3 30B Instruct | 61.7 | #1365 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=personalhomebench-multiple-choice-agentic · How It Works · Data refreshed daily, snapshot 2026-11-21.