Qwen 3.5 2B (Thinking): benchmark results
Provider: Alibaba. Released 2026-03-02. Access: Open.
Aggregate Rating 1437 ± 11, rank #1302 of 2170 rated models, from 108 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LFM3-VL-2B Release - V*Bench | 85.9 | Published normalized score (0–100) | 80 |
| Werewolf Accusation Benchmark - Role Separation | 1.48 | Belief separation (wolves minus villagers, -3..+3 scale) | 79.5 |
| MedLayXPlain | 63.7 | S (self-reported) | 77.4 |
| LFM3-VL-2B Release - HR-Bench (8K) | 74.4 | Published normalized score (0–100) | 70 |
| LFM3-VL-2B Release - RealWorldQA | 75.7 | Published normalized score (0–100) | 70 |
| LFM3-VL-2B Release - DocVQA (val) | 95.6 | Published normalized score (0–100) | 60 |
| LFM3-VL-2B Release - EmbSpatial-Bench | 77 | Published normalized score (0–100) | 60 |
| LFM3-VL-2B Release - HallusionBench | 61.1 | Published normalized score (0–100) | 60 |
| LFM3-VL-2B Release - InfographicVQA (val) | 85.9 | Published normalized score (0–100) | 60 |
| LFM3-VL-2B Release - MLVU (dev, M-Avg) | 68.3 | Published normalized score (0–100) | 60 |
| LFM3-VL-2B Release - MME | 83.9 | Published normalized score (0–100) | 60 |
| LFM3-VL-2B Release - MMMU (val) | 58.1 | Published normalized score (0–100) | 60 |
Interactive version: theaggregate.ai/model?slug=qwen-3-5-2b-thinking · How It Works · Data refreshed daily, snapshot 2026-11-05.