SMH-Bench (Interactive Agent) - Multi-Turn Context: leaderboard

Metric: Success rate (%) on Context-Aware Multi-turn Interaction (TC5, 200 turns that inherit, revise or replace earlier operations); 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

#ModelScore
1Gemini 3.1 Pro (Preview)88.5
2GPT-5.486.5
3Claude Sonnet 4.685
4DeepSeek V3.2 (Thinking)85
5Claude Haiku 4.581
6GPT-5.4 Mini81
7Qwen 3.5 Plus79.5
8GLM-576
9Qwen 3.5 397B A17B74.5
10Qwen 3.5 9B74
11Qwen 3.5 4B65
12MiniMax-M2.741

Interactive version: theaggregate.ai/benchmark?slug=smh-bench-interactive-agent-multi-turn-context · How It Works · Data refreshed daily, snapshot 2026-09-29.