MemLens (128K) - Temporal Reasoning: leaderboard
Metric: Accuracy (%) on the temporal reasoning (TR, 194 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 March 2028. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 235B A22B Instruct | 54.64 |
| 2 | Qwen 3 VL 30B A3B Instruct | 52.58 |
| 3 | Qwen 3 VL 8B Instruct | 44.85 |
| 4 | GLM-4.6V | 43.81 |
| 5 | Qwen 3 VL 235B A22B (Thinking) | 43.3 |
| 6 | Qwen 3.5 122B A10B | 41.75 |
| 7 | Gemini 3.1 Pro (Preview) | 41.24 |
| 8 | GPT-5.4 | 39.69 |
| 9 | Qwen 3 VL 8B (Thinking) | 39.18 |
| 10 | Gemma 3 27B (IT) | 37.11 |
| 11 | Qwen 3 VL 30B A3B (Thinking) | 36.6 |
| 12 | Qwen 3 VL 4B Instruct | 35.05 |
| 13 | Qwen 3.5 27B | 34.02 |
| 14 | Claude Sonnet 4.5 | 32.99 |
| 15 | Phi-4 Multimodal Instruct | 30.93 |
Interactive version: theaggregate.ai/benchmark?slug=memlens-128k-temporal-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.