Alibaba’s Qwen AI lab has just released Qwen 3, a powerful new family of 8 open-weight AI models.
These models come with advanced features like hybrid thinking and smart agent capabilities and they’re already matching the best AI models from OpenAI, DeepSeek, and xAI in performance.
What’s New in Qwen 3?
The star of the show is Qwen3-235B, which performs on par with major models like OpenAI’s o1, Elon Musk’s Grok-3, and DeepSeek-R1.
Qwen3 models now support hybrid thinking model, they can switch between deep reasoning and quick answers. They have improved coding abilities and agentic functions (useful for automation and AI assistants).
The models support a whopping 119 languages. From a lightweight 600M model to the massive 235B, all eight versions are available and show major improvements over earlier releases.
Open for Everyone
All eight models are released under the Apache 2.0 license, meaning developers and companies can use them freely. You can access them on Hugging Face, or deploy them on locally or in the cloud.
Why This Is a Big Deal
Just yesterday, we reported on China closing the gap in advanced AI. Today, Qwen 3 proves that open-source Chinese AI models can now rival the world’s best. With this major launch, all eyes are now on DeepSeek, which is expected to release its R2 model soon.