MemLens (128K) - Multi-Session Reasoning: leaderboard
Metric: Accuracy (%) on the multi-session reasoning (MSR, 143 questions) at a 128K-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: Around 2030. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 29.37 |
| 2 | Kimi K2.5 | 28.67 |
| 3 | GLM-4.6V | 23.08 |
| 4 | Qwen 3.5 122B A10B | 22.38 |
| 5 | Phi-4 Multimodal Instruct | 22.38 |
| 6 | GPT-5.4 | 21.68 |
| 7 | Qwen 3 VL 8B Instruct | 20.28 |
| 8 | Gemma 3 27B (IT) | 18.18 |
| 9 | Gemma 3 12B (IT) | 18.18 |
| 10 | Qwen 3 VL 235B A22B Instruct | 18.18 |
| 11 | Qwen 3 VL 4B Instruct | 16.08 |
| 12 | Qwen 3 VL 30B A3B Instruct | 16.08 |
| 13 | Qwen 3 VL 235B A22B (Thinking) | 15.38 |
| 14 | Qwen 3.5 27B | 13.99 |
| 15 | Gemma 3 4B (IT) | 12.59 |
Interactive version: theaggregate.ai/benchmark?slug=memlens-128k-multi-session-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.