PersonalHomeBench - Feature Recommendation (Agentic): leaderboard

PersonalHomeBench (2026) tests models as assistants in generated, persona-rich smart homes. Feature recommendation ranks appliance features by their relevance to the situation; MAP@1 times 100, acting through the benchmark household toolbox to gather information. Higher is better.

Metric: MAP@1 (times 100) on 1,000 feature recommendation tasks (rank appliance features from API-level inventories by contextual relevance), 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 2029. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-424.7#970
2Qwen 3 30BThinking24.4
3nemotron-3-nano-30B-a3B23.9#1946
4GPT-OSS-20BHigh22.9#1183
5Qwen 3 4BReasoning22.9#2348
6Qwen 3 30B Instruct22.9#1365
7GPT-4o22.7#751

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=personalhomebench-feature-recommendation-agentic · How It Works · Data refreshed daily, snapshot 2026-11-21.