LUNAR - Full Context: leaderboard
Metric: Personalization score (1-5; mean of coverage and depth, GPT-5.1 judge). Source: arxiv.org. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 3.9 |
| 2 | Kimi K2.6 | 3.83 |
| 3 | DeepSeek V4 Flash | 3.57 |
| 4 | GLM-5.1 | 3.44 |
| 5 | MiniMax-M2.7 | 3.32 |
| 6 | DeepSeek V4 Pro | 3.24 |
| 7 | GPT-4.1 Mini | 3.11 |
| 8 | Ling-2.6-1T | 3.06 |
| 9 | Qwen 3 30B A3B 2507 Instruct | 2.91 |
| 10 | GLM-4.7 Flash | 2.85 |
| 11 | GPT-4o Mini | 2.76 |
| 12 | Ling-2.6-flash | 2.76 |
| 13 | Qwen 3 14B (Non-reasoning) | 2.6 |
| 14 | Qwen 3 Next 80B A3B Instruct | 2.58 |
| 15 | Qwen 3 8B (Non-reasoning) | 2.56 |
Interactive version: theaggregate.ai/benchmark?slug=lunar-full-context · How It Works · Data refreshed daily, snapshot 2026-09-19.