Artificial Intelligence: Microsoft, Google and Perplexity integrate the Deepseek R1 model into its platforms

Technological sector companies such as Microsoft, Google and Perplexity have added the model of Artificial intelligence (AI) Deepseek R1 on its platforms and services, in order to bring the reasoning capabilities of this model to more users and companies.

The new R1 reasoning models of the China Depseek technology have generated a great impact on the industry, causing positive responses by companies focused on AI technologies, such as Nvidia and OpenAi, which have valued the progress in the models.

In this context, Microsoft has announced the integration of the Deepseek R1 reasoning model into its Ai Foundry and Github platform, in order to bring their advanced the AI ​​capabilities closer to customers and companies that use these services.

As explained in a statement on its website, Depseek R1 has been added to the catalog of models of both platforms, adding to a “diverse” portfolio that has more than 1,800 models, including vanguard and open source models , as is the case of Deepseek.

In this way, Ai Foundry users will be able to use Deepseek R1 and integrate their abilities in their workflows quickly and easily. All this, from a “reliable, scalable and ready for companies.”

In addition, Microsoft has submitted Deepseek R1 to safety and red equipment evaluations, including “automated evaluations of the model behavior and extensive security reviews to mitigate the possible risks”, as explained by the corporate vice president of the Microsoft AI platform , Asha Sharma.

Following this line, other AI companies, as is the case of Google with its VERTEX AI platform, have also added the R1 model to their services. As explained by the generative product director of Google, Colby Hawker, in an X publication, the model can already be used in vertexai, implementing it directly from Hugging Face.

In the same way, Deepseek R1 has also been implemented in Perplexity services. In this case, the R1 reasoning model is available both in its application and in the web version and allows 500 daily consultations for users signed to the PRO version, as well as 5 daily consultations for the free version.

As shared by the CEO of Technological, Aravind Srinivas, in an X publication, it is the “most powerful” reasoning model in the world and, when integrating into perplexity, has stressed that it is capable of offering “without censorship” results.

Meta plans to add Deepseek’s advances

Meta has also shared his intention to add Depseek’s “advances” in his systems, during his call with investors on the results of the company’s quarter, which have been published on Wednesday.

Specifically, the Meta CEO, Mark Zuckerberg, has detailed that there are some novelties that have carried out since Depseek that are still “assimilating.” He has also advanced that they hope to implement some of these “advances” in their own systems, although it has not deepened in what developments they intend to include. “It is part of the nature of how this works, whether it is a Chinese competitor or not,” he said.

However, regarding infrastructure investment to boost IA, which has been rethink because Depseek has used a much lower and less expensive infrastructure to train its models, Zuckerberg has stated that he advocates to continue investing because “it will be a strategic advantage over time. ”