MemLens (32K) - Multi-Session Reasoning: leaderboard
Metric: Accuracy (%) on the multi-session reasoning (MSR, 143 questions) at a 32K-token history, interleaved multimodal multi-session conversation history with the evidence sessions hidden among haystack sessions, judged correct or incorrect by a Qwen3-VL-235B-A22B-Instruct judge; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K2.5 | 44.06 |
| 2 | Gemini 3.1 Pro (Preview) | 32.17 |
| 3 | Qwen 3.5 122B A10B | 30.07 |
| 4 | Qwen 3.5 27B | 29.37 |
| 5 | Qwen 3 VL 235B A22B (Thinking) | 25.17 |
| 6 | Gemma 3 27B (IT) | 23.78 |
| 7 | Gemma 3 12B (IT) | 22.38 |
| 8 | Gemma 3 4B (IT) | 21.68 |
| 9 | Qwen 3 VL 4B (Thinking) | 20.98 |
| 10 | Phi-4 Multimodal Instruct | 20.28 |
| 11 | GLM-4.6V | 20.28 |
| 12 | Qwen 3 VL 235B A22B Instruct | 18.88 |
| 13 | Qwen 3 VL 30B A3B Instruct | 18.88 |
| 14 | Qwen 3 VL 8B Instruct | 17.48 |
| 15 | Qwen 3.5 2B | 17.48 |
Interactive version: theaggregate.ai/benchmark?slug=memlens-32k-multi-session-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.