PersonalHomeBench - Counterfactual Reasoning (Sole Reasoning): leaderboard

PersonalHomeBench (2026) tests models as assistants in generated, persona-rich smart homes. Counterfactual reasoning asks paired questions whose hypothetical change does or does not alter the answer; accuracy in percent, with the full household context in the prompt and no tools. Higher is better.

Metric: Accuracy (%) on 2,000 counterfactual questions, macro-averaged over the variants that change and that keep the ground-truth answer (exact match, else a Gemini 2.5 Pro semantic-equivalence judge), over the text-based split of PersonalHomeBench (1,100 generated, manually reviewed households with persona-based occupants and appliances), temperature 0, in the sole-reasoning setting (no tools; the full household context, occupant profiles and memories are given in the prompt); 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

#ModelSettingScoreOverall rank
1Qwen 3 30BThinking74.53
2GPT-OSS-20BHigh74#1183
3GPT-473.45#970
4GPT-4o71.3#751
5Qwen 3 4BReasoning71#2348
6nemotron-3-nano-30B-a3B70.83#1946
7Qwen 3 30B Instruct63.55#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-counterfactual-reasoning-sole-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-21.