LLM Stats (MAXIFE): leaderboard
Multilingual Agentic eXecution and Instruction Following Evaluation: tests instruction-following across languages in agentic settings.
Metric: Score (%). Source: llm-stats.com. Source: qwen.ai. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Max | 89.2 | #75 |
| 2 | DeepSeek V4 Pro (Reasoning, Max Effort) | 88.9 | #102 (DeepSeek V4 Pro) |
| 3 | Qwen 3.7 Plus | 88.8 | #92 |
| 4 | Qwen 3.5 397B A17B | 88.2 | #146 |
| 5 | Qwen 3.6 Plus | 88.2 | #120 |
| 6 | Qwen 3.5 27B | 88 | #187 |
| 7 | Qwen 3.5 122B A10B | 87.9 | #173 |
| 8 | Kimi K2.6 (Thinking) | 87.7 | #104 (Kimi K2.6) |
| 9 | GLM-5.1 (Thinking) | 87.7 | #97 (GLM-5.1) |
| 10 | Qwen 3.5 35B A3B | 86.6 | #256 |
| 11 | Qwen 3 Max (Thinking) | 84 | #208 (Qwen 3 Max) |
| 12 | Qwen 3.5 9B | 83.4 | #379 |
| 13 | Claude Opus 4.6 (Thinking) | 81.3 | #63 (Claude Opus 4.6) |
| 14 | Qwen 3.5 4B | 78 | #475 |
| 15 | Qwen 3.5 2B | 60.6 | #1148 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-maxife · How It Works · Data refreshed daily, snapshot 2026-10-20.