Introducing Qwen

GitHub Pages · 2 years ago · 7 ● 2 sources

Alibaba released the Qwen series of open-source large language models, consisting of four models ranging from 1.8B to 72B parameters trained on 2-3 trillion tokens with support for 32K token context length. The largest model, Qwen-72B, achieved competitive performance on benchmarks against Llama 2 and GPT-3.5, with the research team focusing on multilingual capabilities, alignment techniques including SFT and RLHF, and tool-use abilities through agent frameworks. The release enables researchers and developers to build specialized AI applications through open-source models with support for function calling, code interpretation, and agent-based configurations.