MAXIFE: leaderboard
Multilingual instruction-following and execution benchmark evaluating how well models satisfy constraints across languages.
Metric: Score (%). Source: llm-stats.com. Status: saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.7 Max | 89.2 |
| 2 | DeepSeek V4 Pro (Reasoning, Max Effort) | 88.9 |
| 3 | Qwen 3.7 Plus | 88.8 |
| 4 | Qwen 3.6 Plus | 88.2 |
| 5 | Qwen 3.5 397B A17B | 88.2 |
| 6 | Qwen 3.5 27B | 88 |
| 7 | Qwen 3.5 122B A10B | 87.9 |
| 8 | Kimi K2.6 (Thinking) | 87.7 |
| 9 | GLM-5.1 (Thinking) | 87.7 |
| 10 | Qwen 3.5 35B A3B | 86.6 |
| 11 | Qwen 3.5 9B | 83.4 |
| 12 | Claude Opus 4.6 (Thinking) | 81.3 |
| 13 | Qwen 3.5 4B | 78 |
| 14 | Qwen 3.5 2B | 60.6 |
| 15 | Qwen 3.5 0.8B | 39.2 |
Interactive version: theaggregate.ai/benchmark?slug=maxife · How It Works · Data refreshed daily, snapshot 2026-09-05.