Qwen 3.5 Flash: benchmark results
Alibaba's fast, low-cost Qwen 3.5 Flash API tier, the hosted Qwen 3.5 35B A3B. Provider: Alibaba. Released 2026-02-16. Access: API.
Aggregate Rating 1686 ± 9, rank #232 of 1656 rated models, from 216 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI for Education Visual Maths - Measurement | 97.3 | Accuracy (%) | 95.9 |
| SnakeBench | 32.4 | TrueSkill Rating | 92.1 |
| MT-JailBench - Crescendo | 28.3 | Attack success rate (%) of the Crescendo multi-turn jailbrea | 90 |
| MT-JailBench - XTeaming | 23.27 | Attack success rate (%) of the XTeaming multi-turn jailbreak | 90 |
| UAV-DualCog - Landmark-Driven Action Decision | 54.5 | Answer accuracy (%; 1,024 questions asking which way the UAV | 88.6 |
| AI for Education Visual Maths - Statistics and Probability | 57.14 | Accuracy (%) | 88.5 |
| ALL Bench LLM | 75 | Average Numeric Benchmark Score (%) | 88.2 |
| EviPathBench - Image-to-Text Matching | 54.36 | Accuracy (%) pooled over all questions (the All part of each | 88.2 |
| AI for Education Pedagogy - Science | 91.8 | Accuracy (%) | 87.1 |
| IKP - T1 | 100 | Accuracy (%) | 86.1 |
| EviPathBench - Text-to-Image Retrieval | 60.05 | Accuracy (%) pooled over all questions (the All part of each | 85.7 |
| AI for Education Pedagogy - Primary | 92.02 | Accuracy (%) | 84.4 |
Interactive version: theaggregate.ai/model?slug=qwen-3-5-flash · How It Works · Data refreshed daily, snapshot 2026-10-20.