Facts about Qwen
- 07
Qwen's visual language models can process images with up to 1 million tokens of resolution detail, enabling analysis of high-definition documents and complex visual content as of 2024.
- 06
Alibaba open-sourced Qwen models under the Qianwen license in 2023, allowing developers to freely use and modify the models for both commercial and non-commercial purposes.
- 05
Alibaba's Qwen models are available through multiple deployment options including cloud APIs, on-premise installations, and local inference, making them accessible across 150+ countries as of 2024.
- 04
Qwen2.5 models released in October 2024 demonstrated improved coding and mathematics performance through enhanced instruction-following compared to previous Qwen generations.
- 03
Qwen's multilingual capabilities span over 100 languages, with particularly strong performance in Chinese, English, and other major Asian languages as of its 2024 releases.
- 02
In November 2024, Qwen released an open-source model supporting 200,000 token context windows, enabling processing of approximately 150,000 words in a single prompt.
- 01
Alibaba's Qwen 72B model achieved 86.4% accuracy on the MMLU benchmark in 2024, outperforming many larger competing language models.