DeepSeek updated its V3 artificial intelligence model this week with claimed upgrades to the model’s reasoning and programming skills, while also updating its open source license.
The model was updated to V3-0324 on Github, while its license was updated to MIT- a popular open source license originating from the Massachusetts Institute of Technology. DeepSeek did not make a formal announcement on the update.
The model performed substantially better than its predecessor, early testing showed, and appeared to be ahead of comparable thinking models, such as ChatGPT’s o3-mini, AI entrepreneur Paul Gauthier said on X.
DeepSeek’s update comes after the release of its R1 model sent waves through global markets in late-January, as the model appeared to match or even surpass the performance of its rivals while using older hardware and a fraction of their budgets.
DeepSeek also spurred increased confidence in China’s AI capabilities, with a host of other Chinese tech majors, such as Baidu Inc (NASDAQ:BIDU), Bytedance, Alibaba Group Holdings Ltd (NYSE:BABA), and Tencent Holdings Ltd (HK:0700), capitalizing on this popularity by releasing new AI models.
Tencent formally released its Hunyuan T1 reasoning model last week, which it claimed rivaled DeepSeek R1 in performance and price.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.