SMH-Bench (Interactive Agent) - Automation Scheduling: leaderboard
Metric: Success rate (%) on Automated Task Scheduling (TC4, 150 time- or state-triggered rules, trigger and action both verified); Environment-Interactive Agent: a ReAct loop that starts from partial observability and queries and controls the HomeEnv simulator through tools under a call budget; 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 | Qwen 3.5 397B A17B | 69.3 |
| 2 | Claude Sonnet 4.6 | 68.7 |
| 3 | Gemini 3.1 Pro (Preview) | 64 |
| 4 | DeepSeek V3.2 (Thinking) | 64 |
| 5 | GPT-5.4 Mini | 50.7 |
| 6 | GPT-5.4 | 49.3 |
| 7 | Claude Haiku 4.5 | 48 |
| 8 | GLM-5 | 42.7 |
| 9 | Qwen 3.5 Plus | 42.7 |
| 10 | MiniMax-M2.7 | 42 |
| 11 | Qwen 3.5 9B | 36 |
| 12 | Qwen 3.5 4B | 21.3 |
Interactive version: theaggregate.ai/benchmark?slug=smh-bench-interactive-agent-automation-scheduling · How It Works · Data refreshed daily, snapshot 2026-09-29.