Alibaba Cloud’s Qwen2 Model Series Leads in LLM Leaderboard

Alibaba Cloud's Qwen2 Model Series Tops LLM Leaderboard

Alibaba Cloud’s latest language model series, Qwen2, has quickly risen to the top of the rankings for open-sourced LLMs due to its enhanced performance and improved safety alignment. The series includes base language models and instruction-tuned language models with sizes ranging from 0.5 to 72 billion parameters, as well as a Mixture-of-Experts model. The Qwen2 models are available on the Open LLM Leaderboard and Alibaba Cloud’s ModelScope for commercial or research purposes.

The Qwen2-72B model outperforms other leading open-source models in 15 benchmarks, showcasing its capabilities in language understanding, generation, multilingual capability, coding, mathematics, and reasoning. It can handle context lengths up to 128K tokens and includes 27 languages in its training, in addition to Chinese and English. The models also boast increased speed and efficiency in model inference.

In terms of responsible AI, the Qwen2 models demonstrate better alignment with human values and show good performance in safety and responsibility. They can handle multilingual unsafe queries related to illegal activities to prevent misuse. The smaller Qwen2-7B model also excels in benchmarks, including coding.

Alibaba Cloud aims to build an open cloud in the AI era to make computing power more inclusive and AI more accessible. Chief Technology Officer Zhou Jingren emphasizes the importance of accessibility and inclusivity in the field of artificial intelligence.

Source link

Source link:

What do you think?

Leave a Reply

GIPHY App Key not set. Please check settings

DeVonta Smith Pushes Back on Notion He's Underpaid

DeVonta Smith refutes claim of being underpaid in NFL #FairPay

Meta's New AI Tools for WhatsApp

Revolutionizing Business Communications: A Game-Changer for Success with #Innovation