StratMem-Bench - Memory Integration Quality: leaderboard
StratMem-Bench (2026) tests strategic memory use in virtual-character conversation: from a pool of past memories, a reply must use the memories the query requires and may use others that enrich it. This board is a judge score from 1 to 5 for how well memories are integrated into replies that pass the coverage check; higher is better.
Metric: Memory integration quality MIQ (1-5, judge-rated) of responses that pass the SMC check, over all instances weighted by scenario counts; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Chat | 4.56 | #449 |
| 2 | GPT-5.2 | 4.45 | #185 |
| 3 | Llama 4 Maverick | 4.44 | #1110 |
| 4 | DeepSeek V3.2 | 4.4 | #419 |
| 5 | Claude Sonnet 4.5 | 4.37 | #261 |
| 6 | Qwen 3 Max | 4.3 | #415 |
| 7 | Qwen 3 235B A22B 2507 Instruct | 4.24 | #616 |
| 8 | Gemini 3 Pro (Preview) | 4.21 | #119 |
| 9 | StratMem DeepSeek-reasoner (checkpoint unspecified) | 4.12 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=stratmem-bench-memory-integration-quality · How It Works · Data refreshed daily, snapshot 2026-11-21.