EverMemBench (Full Context) - Fine-Grained Recall (Temporal): leaderboard
Metric: Accuracy (%; whole multi-party dialogue history in context, open-ended, judged by an LLM). Source: arxiv.org. Saturation forecast: Around December 2026. 3 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 45 |
| 2 | GPT-4.1 Mini | 7 |
| 3 | Llama 4 Scout Instruct | 1.67 |
Interactive version: theaggregate.ai/benchmark?slug=evermembench-full-context-fine-grained-recall-temporal · How It Works · Data refreshed daily, snapshot 2026-09-26.