QwenB

Alibaba Cloud’s open-weight foundation model series: high-performance reasoning for multilingual and multimodal tasks.

QwenB is the foundation model series driving Alibaba Cloud’s AI ecosystem: a collection of high-performance LLMs built for industrial-scale deployment. These models are trained on over 3 trillion tokens and support 100+ languages (topping open-source benchmarks like MMLU and GSM8K). The series features diverse architectures (spanning 1.8B to 72B parameters) and handles complex inputs across text, vision, and audio: specifically through the Omni-7B multimodal variant. With an Apache 2.0 license, QwenB provides the raw weights necessary for developers to build specialized agents that rival proprietary frontier models in reasoning and accuracy.

https://qwen.ai