Qwen3.5-2B-Base: benchmark results
Alibaba's Qwen3.5-2B-Base, a pretrain-only checkpoint in the new Qwen3.5 family intended for fine-tuning and in-context-learning research rather than direct interaction, with vision-language pretraining. Provider: Alibaba. Released 2026-03-02. Access: Open.
Aggregate Rating 1374 ± 23, rank #1288 of 1640 rated models, from 292 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Italian NLU - SQuAD IT | 74.47 | Character F1 (%) | 93.2 |
| EuroEval English NLU - SQuAD | 86.3 | Character F1 (%) | 91.7 |
| EuroEval German NLU - Germanquad | 65.71 | Character F1 (%) | 91.6 |
| EuroEval French NLU - FQuAD | 73.93 | Character F1 (%) | 91.2 |
| EuroEval Greek NLU - MultiWikiQA EL | 68.96 | Character F1 (%) | 89.7 |
| EuroEval Finnish NLU - Tydiqa FI | 72.21 | Character F1 (%) | 89.3 |
| EuroEval Dutch NLU - SQuAD NL | 78.88 | Character F1 (%) | 89.2 |
| EuroEval Spanish NLU - MLQA ES | 66.89 | Character F1 (%) | 89.2 |
| EuroEval Icelandic NLU - NQII | 56.65 | Character F1 (%) | 89 |
| EuroEval Serbian NLU - MultiWikiQA SR | 63.45 | Character F1 (%) | 88.2 |
| EuroEval Albanian NLU - MultiWikiQA SQ | 60.91 | Character F1 (%) | 84.9 |
| EuroEval Estonian NLU - MultiWikiQA ET | 68.59 | Character F1 (%) | 83.9 |
Interactive version: theaggregate.ai/model?slug=qwen3-5-2b-base · How It Works · Data refreshed daily, snapshot 2026-11-07.