Deepseek not only unleashed madness between users and companies that have AI models for being Open Sourcebut for its efficiency so that it can even be installed locally on a mobile. Now the Startup China has updated the reasoning model Deepseek R1.
Another of Deepseek R1’s greatest virtues, apart from being completely free without limits for users, is that shows your logic of thought to reason the user’s responses. This experience can also be enjoyed in Chatgpt as in Gemini.
Now the Startup China hit a big jump forward when publishing an updated version of its R1 Reasoning Model on the Hugging Face developer platform by announcing it from Wechat a few hours ago.
It is a minor update according to Crow Deepseek in the announcement in Wechat, and arrives under a permissive mit license. Which means that It can be commercially used Without limits as another characteristic that opened many possibilities when it arrived in January.
In the repository of Hugging Face No description of the model appears, according to Techcrunchand it is left alone in the configuration files and the internal components of a model that guides its behavior, but in Reddit some of its best capabilities have already been shared, which are not few.
The highest performance AI models from LiveCodebench
Deepseek R1-0528 already reason deeply As Google models, their capacity for creative writing has been improved to be more natural and with better format, and now uses a different reasoning style focused on being more reflective than fast.
There is another important point in the Deepseek R1 update: it is able to generate deep -thoughts of greater extension for Reach up to 30 – 60 minutesso consultations may even be more complex.
As for performance, the new Deepseek R1-0528 yields along with OPENAI O3 according to LivecodeBenchthe most advanced chatgpt model that was released last month. So he becomes a total challenge for OpenAi’s interests with his chatgpt.
With regard to its availability, the new Deepseek R1-0528 updated model can already be tested in the mobile apps and on the official websiteso you can access to receive answers that can now exceed more than 2,000 words (before staying in the thousand).
To be able to install it locally, according to Unslothyour download would be available in the next 24 hours. This has been another of its best capabilities, to be able to install it on a PC provided that there is an RTX card of the 30xx, 40xx series or the last 50xx (although it is not entirely essential to make it work).
Since Deepseek R1 was updated, they have already consumed more than 500 million tokensa very high figure for an AI model that does not have even 24 hours available. Be Open Sourcewithout limits, that operates as a gemini and reaches O3 again makes clear the ability of the Startup China.
You are still waiting for you to launch Deepseek R2which would be a major step in capacities and performance. It was supposed to be last month when the new version would arrive, but it seems that we will have to wait and enjoy this new update that puts Openai’s chatgpt against the ropes again.
