Qwen 3.5 2B: benchmark results
2B on-device member of Alibaba's Apache-2.0 Qwen3.5 family (March 2026), sized for phones and laptops. Provider: Alibaba. Released 2026-03-02. Access: Open.
Aggregate Rating 1420 ± 7, rank #1148 of 1656 rated models, from 907 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Bulgarian NLU - Cinexio | 58.3 | Matthews correlation coefficient (%) | 93.4 |
| Kurakurai French LLM - GPQA Diamond | 36.8 | Accuracy (%) | 92.7 |
| UnpredictaBench - Shuffling | 36.84 | KS@100 (%): share of the 19 Shuffling problems (random order | 92.1 |
| Kurakurai French LLM - MMLU-Pro | 47.2 | Accuracy (%) | 90.2 |
| SAHARA - Paraphrase | 32.08 | spBLEU-1K | 89.4 |
| EuroEval French NLU - ScaLA FR | 52.57 | Matthews correlation coefficient (%) | 88 |
| Kurakurai French LLM - AIME 2024 | 11.67 | Accuracy (%) | 87.8 |
| EuroEval Portuguese NLU - ScaLA PT | 35.02 | Matthews correlation coefficient (%) | 87.3 |
| EuroEval Italian NLU - ScaLA IT | 40.14 | Matthews correlation coefficient (%) | 86.6 |
| EuroEval Catalan NLU - Guia CAT | 68.8 | Matthews correlation coefficient (%) | 85.8 |
| EuroEval Albanian Summarization - LR SUM SQ | 33.48 | chrF3++ (%) | 83.7 |
| EuroEval Catalan NLU - MultiWikiQA CA | 72.51 | Character F1 (%) | 81.5 |
Interactive version: theaggregate.ai/model?slug=qwen-3-5-2b · How It Works · Data refreshed daily, snapshot 2026-10-20.