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
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 78 |
| 2 | Qwen 3.5 397B A17B | 76.7 |
| 3 | Gemini 3.1 Pro (Preview) | 72.7 |
| 4 | Claude Haiku 4.5 | 72.7 |
| 5 | DeepSeek V3.2 (Thinking) | 65.3 |
| 6 | Claude Sonnet 4.6 | 61.3 |
| 7 | GPT-5.4 Mini | 61.3 |
| 8 | MiniMax-M2.7 | 60.7 |
| 9 | Qwen 3.5 Plus | 51.3 |
| 10 | GLM-5 | 49.3 |
| 11 | Qwen 3.5 9B | 47.3 |
| 12 | Qwen 3.5 4B | 44.7 |
Interactive version: theaggregate.ai/benchmark?slug=smh-bench-direct-reasoning-ambiguous-intent · How It Works · Data refreshed daily, snapshot 2026-09-29.