AMemGym: leaderboard
Metric: Normalized memory score (times 100): overall question-answering accuracy rescaled between a random-guess baseline (0) and the same model's upper bound when given the ground-truth user states (100), averaged over evaluation periods, on the AMemGym base configuration: 20 synthetic user profiles, each with 10 multiple-choice personalization questions (4 to 7 options) asked after each of 11 interaction periods while an LLM-simulated user (GPT-4.1) reveals evolving user states over about 47 turns; the native LLM keeps the whole history in context; temperature 0; on-policy: the model's own conversation with the simulated user is the history; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4 | 33.6 | #194 |
| 2 | Gemini 2.5 Flash | 32.7 | #237 |
| 3 | Gemini 2.5 Flash Lite | 26.9 | #413 |
| 4 | GLM-4.6 | 25.7 | #246 |
| 5 | GPT-4.1 | 24.4 | #240 |
| 6 | Gemini 2.0 Flash | 24.4 | #331 |
| 7 | Kimi K2 | 23.4 | #236 |
| 8 | GPT-4.1 Mini | 20.3 | #346 |
| 9 | DeepSeek V3 | 15.2 | #312 |
| 10 | DeepSeek V3.1 Terminus | 15.1 | #212 |
| 11 | GPT-4o Mini | 14.9 | #588 |
| 12 | Qwen 3 235B A22B 2507 Instruct | 14.8 | #291 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=amemgym · How It Works · Data refreshed daily, snapshot 2026-10-11.