Qwen 3 4B Instruct: benchmark results
Alibaba's open 4B dense model from the original April 2025 Qwen3 release, used in instruct (non-thinking) mode. Provider: Alibaba. Released 2025-04-29. Access: Open.
Aggregate Rating 1535 ± 14, rank #657 of 1656 rated models, from 88 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| IBA-Bench - Medical Services | 82.9 | Checkpoint satisfaction (%; share of hidden preference check | 100 |
| IBA-Bench - Transportation | 80 | Checkpoint satisfaction (%; share of hidden preference check | 100 |
| Fibble4 Arena | 6.67 | Daily clean-run win rate (%), 12 guesses | 88.9 |
| IBA-Bench - Leisure Activities | 52.6 | Checkpoint satisfaction (%; share of hidden preference check | 88.9 |
| SWE-QA (Multi-Hop Code) - Declaration-and-Call | 74.58 | Accuracy (%) on the 4,584 Declaration-and-Call questions, wh | 84.6 |
| SWE-QA (Multi-Hop Code) - Interacting Entities | 65.41 | Accuracy (%) on the 4,488 Interacting-Entity questions, whic | 84.6 |
| EComStage - Solution Decision | 84.8 | Task Score (%) | 81.2 |
| EComStage | 82.26 | Task Score (%) | 68.8 |
| EComStage - Attitude Classification | 82.78 | Task Score (%) | 68.8 |
| EComStage - Query Rewrite | 79.97 | Task Score (%) | 68.8 |
| IBA-Bench - Daily Consumption | 77.8 | Checkpoint satisfaction (%; share of hidden preference check | 66.7 |
| IBA-Bench - Exercise and Health | 73.4 | Checkpoint satisfaction (%; share of hidden preference check | 66.7 |
Interactive version: theaggregate.ai/model?slug=qwen-3-4b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.