Alibaba has launched Qwen 3, a family of AI models that the company claims can compete with and sometimes exceed the performance of top models from Google and OpenAI. These models, which vary in size from 0.6 billion to 235 billion parameters, are available for download via Hugging Face and GitHub. With advancements enabling reasoning and fact-checking, Qwen 3 supports 119 languages and was trained on an extensive set of 36 trillion tokens, significantly enhancing its performance over its predecessor, Qwen 2.
The Qwen 3 models are hybrid models that enable both reasoning and quick responses, providing users with control over the processing speed.
With a training set of nearly 36 trillion tokens and capabilities in 119 languages, Qwen 3 surpasses previous models offering improved performance.
Collection
[
|
...
]