STALE: leaderboard
Large Language Model (LLM) agents are increasingly expected to maintain coherent, long-term personalized memory, yet current benchmarks primarily measure static fact retrieval, overlooking the ability to revise stored beliefs when new evidence emerges. We identify a critical and underexplored failure mode, Implicit Conflict: a later observation invalidates an earlier memory without explicit negati
Metric: Overall (self-reported). Source: benchmarklist.com. Saturation forecast: Around June 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 55.2 | #58 |
| 2 | Qwen 3.5 27B | 31.3 | #187 |
| 3 | Gemini 3.1 Flash Lite | 22.4 | #191 |
| 4 | GPT-5.4 | 15.7 | #79 |
| 5 | Qwen 3.5 9B | 14.6 | #379 |
| 6 | GPT-5.4 Nano | 9.8 | #328 |
| 7 | GPT-4o Mini | 8.7 | #599 |
| 8 | MiniMax-M2.5 | 5.5 | #303 |
| 9 | Llama 3.3 70B Instruct | 2.6 | #534 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=stale · How It Works · Data refreshed daily, snapshot 2026-10-20.