SMH-Bench (Direct Reasoning) - Atomic Control: leaderboard

Metric: Success rate (%) on Atomic Control (TC1, 150 single-intent device operations under clear, colloquial or noisy wording); 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: Estimated already saturated. 13 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)95.3
2Qwen 3.5 397B A17B92.7
3GPT-5.491.3
4Claude Sonnet 4.690.7
5Claude Haiku 4.590
6MiniMax-M2.781.3
7GPT-5.4 Mini80.7
8GLM-580.7
9Qwen 3.5 Plus79.3
10DeepSeek V3.2 (Thinking)79.3
11Qwen 3.5 9B73.3
12Qwen 3.5 4B68.7

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