SMH-Bench (Direct Reasoning) - Ambiguous Intent: leaderboard

Metric: Success rate (%) on Ambiguous Intent (TC3, 150 subjective complaints, scene-level goals and slot-missing requests, acting or asking for clarification as appropriate); Direct Reasoning: the full serialized home state (rooms, devices, current values, service schemas) and the request in one prompt, one structured JSON answer replayed in the HomeEnv simulator; final home state checked against the goal with untouched devices preserved; clarification and query answers judged by GPT-5; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 13 models tracked.

Top models

#ModelScore
1GPT-5.478
2Qwen 3.5 397B A17B76.7
3Gemini 3.1 Pro (Preview)72.7
4Claude Haiku 4.572.7
5DeepSeek V3.2 (Thinking)65.3
6Claude Sonnet 4.661.3
7GPT-5.4 Mini61.3
8MiniMax-M2.760.7
9Qwen 3.5 Plus51.3
10GLM-549.3
11Qwen 3.5 9B47.3
12Qwen 3.5 4B44.7

Interactive version: theaggregate.ai/benchmark?slug=smh-bench-direct-reasoning-ambiguous-intent · How It Works · Data refreshed daily, snapshot 2026-09-29.