Alibaba has unveiled its latest artificial intelligence model, Qwen3, boasting superior test results in certain aspects compared to DeepSeek-R1, o1, o3-mini, Grok-3, and Gemini-2.5-Pro. Qwen3 comes in two sub-models: Qwen3-235B-A22B with total parameters of 235B and operational parameters of 22B, and the smaller model Qwen3-30B-A3B with total parameters of 30B and operational parameters of 3B. Additionally, there are 6 more open-weighted models ranging from 0.6B to 32B.
Alibaba highlights Qwen3’s hybrid working mode, which includes Think Mode for step-by-step processing and No-Think Mode for speedy responses to simple questions. Users can adjust the model’s thinking levels based on their needs and processing costs. Moreover, Qwen3 supports up to 119 languages, including Thai.
Qwen3 is an open-source model available for further exploration on Hugging Face and GitHub.
TLDR: Alibaba introduces Qwen3, a cutting-edge AI model with superior performance and customizable thinking levels, supporting 119 languages and available as an open-source model on Hugging Face and GitHub.
Leave a Comment