According to news on December 8, HuggingFace, the world's largest open source large model community, recently announced the latest open source large model rankings. Alibaba Cloud Tongyi Qianwen topped the list of domestic and foreign open source large models such as Llama2.
HuggingFace's Open LLM Leaderboard is currently the most authoritative list in the field of large models. It includes hundreds of open source large models from around the world. The test dimensions cover reading comprehension, logical reasoning, mathematical calculations, fact answering, etc. Six major reviews. Tongyi Qianwen (Qwen-72B) performed well, ranking first among all pre-trained models with a comprehensive score of 73.6.
Tongyi Qianwen-72B climbed to the top of HuggingFace’s Open LLM Leaderboard
In early December, Alibaba Cloud announced that it would officially open source the 72 billion-parameter large language model Tongyi Qianwen Qwen-72B. Qwen-72B achieved the best results among open source models in 10 authoritative benchmarks and became the industry's most powerful open source large model, with performance exceeding that of open source. The benchmark Llama 2-70B and most commercial closed-source models can be adapted to high-performance applications at the enterprise level and scientific research level.
Alibaba Cloud is the first technology company in China to open source large self-developed models. Since August this year, it has successively open sourced Qwen-7B, Qwen-14B, Qwen-1.8B, visual understanding model Qwen-VL, and audio understanding large model Qwen-Audio, taking the lead in Realized the "full size, full mode" open source of large models. Several models have successively appeared on the HuggingFace and Github large model lists, and are widely favored by small and medium-sized enterprises and individual developers. The cumulative downloads have exceeded 1.5 million, and more than 150 new models and new applications have been spawned.