Qwen 3.5 0.8B (Non-reasoning): benchmark results
Provider: Alibaba. Released 2026-03-02. Access: Open.
Aggregate Rating 1273 ± 16, rank #1912 of 2170 rated models, from 78 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PRISM-VLM - Sycophancy Resistance | 81 | Per-axis mean score (%; share of turn-0-correct answers held | 97.6 |
| Tau2-Bench Telecom | 65.2 | Success Rate (%) | 60.5 |
| AA TAU-2 Bench | 65.2 | Accuracy (%) | 59.2 |
| ToolUse.1 - Abstention | 0.84 | Rate (0–1) | 50 |
| ToolUse.1 - Spurious Call Rate | 0.06 | Rate (0–1) | 50 |
| ToolUse.1 - XML | 0.68 | Rate (0–1) | 50 |
| PRISM-VLM - Multi-Question Robustness | 29.8 | Per-axis mean score (%; share of items whose 3-5 bundled que | 31.7 |
| AA Omniscience - Software Engineering (SWE) - Julia | 4 | Accuracy (%) | 30.4 |
| AA Humanity's Last Exam | 5.1 | Accuracy (%) | 28.4 |
| PRISM-VLM | 45.1 | PScore (%; harmonic mean of the seven per-axis mean scores, | 27.5 |
| UGI - Willingness (W/10) | 3.5 | W/10 Score | 27.2 |
| ToolUse.1 - JSON | 0.68 | Rate (0–1) | 25 |
Interactive version: theaggregate.ai/model?slug=qwen-3-5-0-8b-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.