Hy-MultiTurn - Constraint Memory: leaderboard
Metric: Importance-weighted score (%) on the 34 constraint memory dialogues, 209 long Chinese multi-turn dialogues (12 to 76 turns), three rolls per task averaged, three LLM judges (GPT-5.2, Qwen 3.6, DeepSeek-V4 Pro) averaged, a deterministic checker for precise execution; higher is better. Source: arxiv.org. Saturation forecast: Around July 2027. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 (High) | 71.2 |
| 2 | Gemini 3.1 Pro (Preview) (High) | 68 |
| 3 | GPT-5.4 (High) | 61.8 |
| 4 | Claude Opus 4.6 (High) | 59.1 |
| 5 | Kimi K3 (Max) | 58 |
| 6 | Kimi K2.6 | 56.3 |
| 7 | Hy3 (High) | 55.4 |
| 8 | Claude Opus 4.7 (High) | 52.2 |
| 9 | Grok 4.5 (High) | 48.9 |
| 10 | Kimi K2.5 (Thinking) | 48.2 |
| 11 | Hy3-preview (High) | 43.8 |
| 12 | Qwen 3.6 Plus (Thinking) | 41.8 |
| 13 | Doubao-Seed-2.1-Pro (High) | 40.8 |
| 14 | Kimi K2.6 (Non-reasoning) | 35.3 |
| 15 | Gemini 3.1 Flash Lite (Non-reasoning) | 34.4 |
Interactive version: theaggregate.ai/benchmark?slug=hy-multiturn-constraint-memory · How It Works · Data refreshed daily, snapshot 2026-09-29.