MemLens (64K) - Temporal Reasoning: leaderboard
Metric: Accuracy (%) on the temporal reasoning (TR, 194 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 2028. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 30B A3B Instruct | 55.15 |
| 2 | Qwen 3 VL 235B A22B Instruct | 54.12 |
| 3 | GLM-4.6V | 53.09 |
| 4 | Kimi K2.5 | 52.06 |
| 5 | Qwen 3.5 122B A10B | 51.55 |
| 6 | Qwen 3 VL 8B (Thinking) | 50.52 |
| 7 | Qwen 3 VL 8B Instruct | 47.42 |
| 8 | Gemma 3 27B (IT) | 44.33 |
| 9 | Gemma 3 12B (IT) | 44.33 |
| 10 | Qwen 3 VL 235B A22B (Thinking) | 42.78 |
| 11 | Nemotron Nano 12B v2 VL | 42.78 |
| 12 | Gemini 3.1 Pro (Preview) | 42.27 |
| 13 | GPT-5.4 | 42.27 |
| 14 | Phi-4 Multimodal Instruct | 39.18 |
| 15 | Qwen 3 VL 30B A3B (Thinking) | 37.11 |
Interactive version: theaggregate.ai/benchmark?slug=memlens-64k-temporal-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.