Aliyun announces open source of Tongyi Qianwen 72 billion parameter model

CTOnews.com news on December 1, Aliyun announced the open source Thousand ask 72 billion parameter model Qwen-72B, 1.8 billion parameter model Qwen-1.8B and audio large model Qwen-Audio, which has been online.

In this open source model, in addition to the pre-training model, the corresponding dialogue model is also introduced simultaneously, and a quantitative version of 4bit / 8bit model is provided for 72B and 1.8B dialogue model, which is convenient for developers to reason and train.

CTOnews.com Note: Aliyun has previously opened up 7 billion parameter model Qwen-7B and 14 billion parameter model Qwen-14B.

According to reports, Qwen-72B is equipped with a strong system instruction (System Prompt) capability, and users can customize their AI assistant with only one prompt, requiring the big model to play a role or perform specific reply tasks.

According to officials, Qwen-72B achieved the best results of the open source model in 10 authoritative benchmark tests, making it the strongest open source model in the industry, surpassing the open source benchmark Llama 2-70B and most commercial closed source models.

In the future, enterprise-level and scientific research-level high-performance applications will also have the option of open source large model. Tongyi Qianwen also opened up 1.8 billion parameter model Qwen-1.8B and audio large model Qwen-Audio.

So far, Tongyi Qianqian has opened up four large language models with 1.8 billion, 7 billion, 14 billion and 72 billion parameters, as well as two multimodal models for visual understanding and audio understanding, realizing "full-size, full-mode" open source.

CTOnews.com with a model link:

General meaning thousand questions-72B-pre-training:

Https://modelscope.cn/models/qwen/Qwen-72B

General meaning Thousand questions-72B-Chat:

Https://modelscope.cn/models/qwen/Qwen-72B-Chat

General meaning Thousand questions-72B-Chat-Int8:

Https://www.modelscope.cn/models/qwen/Qwen-72B-Chat-Int8

General meaning Thousand questions-72B-Chat-Int4:

Https://www.modelscope.cn/models/qwen/Qwen-72B-Chat-Int4

Tongyi thousand questions-1.8B-pre-training:

Https://modelscope.cn/models/qwen/Qwen-1_8B

General meaning Thousand questions-1.8B-Chat:

Https://modelscope.cn/models/qwen/Qwen-1_8B-Chat

General meaning Thousand questions-1_8B-Chat-Int8:

Https://www.modelscope.cn/models/qwen/Qwen-1_8B-Chat-Int8

General meaning Thousand questions-1_8B-Chat-Int4:

Https://www.modelscope.cn/models/qwen/Qwen-1_8B-Chat-Int4

General meaning Thousand questions-Audio- pre-training:

Https://modelscope.cn/models/qwen/Qwen-Audio

General meaning Thousand questions-Audio-Chat:

Https://modelscope.cn/models/qwen/Qwen-Audio-Chat

Download the model (take Tongyi Qianwen-72B-Chat-Int4 as an example):

From modelscope import snapshot_downloadmodel_dir = snapshot_download ("qwen/Qwen-72B-Chat-Int4")