History
Qwen3 offers models in a wide range of sizes
The Qwen3 family arrived with available weights, an Apache 2.0 licence and different thinking modes.

On 29 April 2025, the Qwen team introduced Qwen3. The family ranged from small to large models, with available weights under Apache 2.0. The models combined a mode with extra reasoning effort and a mode without that additional work. The announcement describes the family.

In our view, that breadth made the moment relevant. A choice concerned more than one large model: it also involved the size and approach a task required. The family name does not make that decision for you.

For a small or medium-sized business, a trial could involve two bounded assignments. Have short public questions sorted into categories, and separately have a practice schedule checked. Specify beforehand what a good answer looks like for each task. Then compare suitable model sizes and settings using the same examples. Measure the required equipment and review effort too. Ask your builder to record the chosen version and usage terms. This assesses a model family through the components you actually use, leaving room to carry out a simple task in a simple way.