Alibaba Launches Qwen3, a New Benchmark in Open-Source AI

Alibaba has introduced Qwen3, the latest in its open-sourced large language model series, setting a new standard for AI innovation. Featuring six dense and two Mixture-of-Experts models, Qwen3 offers developers a flexible toolkit for next-generation applications across various platforms, including mobile devices and autonomous vehicles.

The Qwen3 models are globally available as open-source tools. This new series marks Alibaba's debut into hybrid reasoning models, integrating traditional capabilities with advanced dynamic reasoning. Developers can utilize the API to control thinking duration for an optimized balance of performance and efficiency.

Trained on extensive datasets, Qwen3 excels in multilingual skills, agent capabilities, and reasoning, outperforming its predecessors. It boasts superior results in industry benchmarks, making significant strides in mathematics, coding, and multilingual tasks.

Qwen3 is accessible on Hugging Face, GitHub, and ModelScope, with API access forthcoming via Alibaba's Model Studio. Since launch, it has seen significant adoption, reinforcing its place as a leading open-source AI model series.