Post
2840
Qwen3 is the latest version of the Qwen language models. It's smarter, faster, and now understands 119 languages instead of just 29.
It can do both deep reasoning and quick answers using a single model, depending on what you need.
The models range in size from small (0.6B) to huge (235B), with smart ways to save compute.
It's trained on 36 trillion tokens and fine-tuned in four steps to boost performance.
Qwen3 performs as well as or better than many top models, including some from big companies.
It’s fully open-source under licence. Amazing!!!
https://github.com/QwenLM/Qwen3/blob/main/Qwen3_Technical_Report.pdf
It can do both deep reasoning and quick answers using a single model, depending on what you need.
The models range in size from small (0.6B) to huge (235B), with smart ways to save compute.
It's trained on 36 trillion tokens and fine-tuned in four steps to boost performance.
Qwen3 performs as well as or better than many top models, including some from big companies.
It’s fully open-source under licence. Amazing!!!
https://github.com/QwenLM/Qwen3/blob/main/Qwen3_Technical_Report.pdf