Qwin 2

Qwen2 is Alibaba Cloud's next-generation series of open-source large language models featuring enhanced multilingual capabilities and state-of-the-art performance in coding and mathematics.

Alibaba Cloud's Qwen2 series represents a major leap in open-source AI, offering five distinct model sizes ranging from 0.5B to 72B parameters. These models are pretrained on over 27 languages and utilize Group Query Attention (GQA) to optimize inference speed and memory efficiency. With a massive context window of up to 128K tokens, Qwen2 excels at processing complex datasets and long-form content. The flagship 72B-Instruct model consistently tops leaderboards (outperforming Llama-3-70B in key benchmarks) while maintaining strict alignment with human values through advanced post-training techniques.

https://qwenlm.github.io/blog/qwen2/