MemLens (64K) - Information Extraction: leaderboard
Accuracy in percent on the MemLens 246 questions that extract a fact or image detail from one earlier session, at a 64K-token interleaved image and text conversation history, with the evidence sessions hidden among distractor sessions (MemLens, 2026), judged by Qwen3-VL-235B. Higher is better.
Metric: Accuracy (%) on the information extraction (IE, 246 questions) at a 64K-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 February 2027. 27 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 122B A10B | 67.07 | #350 |
| 2 | GPT-5.4 | 63.01 | #138 |
| 3 | Qwen 3.5 27B | 62.2 | #380 |
| 4 | GLM-4.6V | 61.79 | #721 |
| 5 | Qwen 3.5 9B | 60.57 | #838 |
| 6 | Gemini 3.1 Pro (Preview) | 58.94 | #102 |
| 7 | Qwen 3 VL 30B A3B Instruct | 56.91 | #903 |
| 8 | Qwen 3 VL 235B A22B Instruct | 56.1 | #541 |
| 9 | Nemotron Nano 12B v2 VL | 56.1 | #2385 |
| 10 | Kimi K2.5 | 52.44 | #272 |
| 11 | Qwen 3.5 4B | 49.19 | #1270 |
| 12 | Cosmos-Reason2-8B | 46.34 | #758 |
| 13 | GLM-4.5V | 45.93 | #756 |
| 14 | Qwen 3 VL 8B Instruct | 44.72 | #1043 |
| 15 | Qwen 3 VL 4B Instruct | 42.28 | #1348 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=memlens-64k-information-extraction · How It Works · Data refreshed daily, snapshot 2026-11-22.