Alibaba Introduces Qwen3, Open-source AI Model With Hybrid Reasoning

Alibaba launched Qwen3 (photo: dock. unsplash)

JAKARTA Alibaba launched the latest Artificial Intelligence (AI) model named Qwen3. This AI model is the latest generation of a large language model (LLM) that they introduced last April.

Qwen3 consists of six model dense and two Mixture-of-Expert (MoE) models that can provide flexibility to developers. All Qwen3 models have been launched in open-source form and are globally accessible.

This latest AI model is an achievement in the hybrid reasoning model. Alibaba combines Thought Mode and Non-Thinking which combines traditional LLM capabilities with advanced dynamic reasoning so that Qwen3's work feels smooth.

When a user uses thinking mode, this model can carry out complex tasks and requires several solution steps such as math or coding. Meanwhile, mode without thinking is used to get a common response.

For developers, Qwen3 can be accessed through the API and this AI model can control the thinking duration of up to 38,000 tokens. That way, they can maintain a balance between intelligent performance and computer efficiency.

If you want to reduce the cost of implementing significantly compared to other AI models, developers can use the MoE Qwen3-235B-A22B. This model is proof that Alibaba is committed to developing high-performance AI with easy access.

In addition to some of the capabilities above, Qwen3 also offers several other capabilities. This AI model supports 119 languages and high-performance dialects for translational needs and supports the Context Protocol (MCP) Model natively.

Qwen3 can also be used to improve writing experience or other creative needs. For example, users can change the language of dialogue in theater works with a more interesting or more natural closure.