HEART-Bench: leaderboard
Metric: Accuracy (%) averaged over the three memory frameworks (Naive-RAG, Mem0 and PersonaDB), times 100, 673 four-option multiple-choice questions on what one of 11 Big Five-grounded characters would do in a scenario, the prompt exposing only the character ID, occupation and retrieved de-identified episodic memories (Qwen3-Embedding-4B retriever, GPT-5.4-mini auxiliary memory manager); higher is better. Source: arxiv.org. Saturation forecast: Around April 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 62.7 |
| 2 | Gemini 3 Flash | 54.3 |
| 3 | DeepSeek V3.2 | 42.2 |
| 4 | Qwen 3.5 397B A17B | 40.4 |
| 5 | Claude Sonnet 4.6 | 39.7 |
| 6 | Qwen 3.5 122B A10B | 39.1 |
| 7 | DeepSeek V4 Pro | 38.2 |
| 8 | Qwen 3.5 35B A3B | 36.9 |
| 9 | Claude Haiku 4.5 | 36.5 |
| 10 | GPT-5.4 | 36.3 |
| 11 | DeepSeek V4 Flash | 33.7 |
| 12 | GPT-5.4 Mini | 33.2 |
Interactive version: theaggregate.ai/benchmark?slug=heart-bench · How It Works · Data refreshed daily, snapshot 2026-10-07.