DeepSeek, a Chinese startup, has recently launched an upgraded version of its R1 AI reasoning model this Wednesday morning on WeChat. The new model, known as R1-0528, will launch on the Hugging Face developer platform. Moreover, there’s no official announcement made regarding this publicly.
DeepSeek’s R1 AI Model: A Solid Competitor
The new version stands as a strategic competition against the big AI players, OpenAI. DeepSeek has gained popularity in the global markets after its open-source and free R1 reasoning model from competitors in the industry. The R1-058 model currently is in line behind OpenAI’s o4-mini and o3 models when tested on the LiveCode leaderboard, a site that analyzes the reasoning capabilities of AI models.
During the announcement, the company stated that the R1-058 model is just a slight upgrade. However, with 685 billion parameters, the R1 is quite heavy in weight, and cannot run on most consumer-grade hardware.
Alongside, it’s licensed under the permissive MIT license with access to commercial use. The Hugging Face repository comes with configuration files and weights, and no complete description.
Further, the announcements are here with rising international tensions and increased American concern over China’s staying one step ahead in AI developments.
The rapid development of DeepSeek highlights China’s constant work in AI research despite sanctions and points to a change in the global balance of AI innovation.
Additionally, the business is looking to introduce R2, the replacement for R1. In May, the release was anticipated. DeepSeek also improved its V3 big language models in March with this.
To stay in the know of such trending news, visit us now!
Recommended For You: