- The startup of Chinese AI Deepseek published an improved AI model called V3-0324 to hug the face
- V3-0324 offers improved reasoning and coding capacities on its predecessors
- Deepseek claims that its AI models can correspond or beat those of American AI developers like Openai and Anthropic
Deepseek abandoned a major upgrade of his IA model this week, which makes people buzz almost as much as they did when the Chinese AI startup has splashed earlier this year. The new Deepseek-V3-0324 model is now live on an embraced face, setting up an even more contributed rivalry with OPENAI and other IA developers.
According to the company’s tests, the new Deepseek iteration of its V3 model offers measurable increases in reasoning and coding capacity. Better reflection and coding might not seem revolutionary by themselves, but the rate of improvement and the Deepseek plans make this version notable.
Trained last year, Deepseek evolved quickly, starting with the release in December of the original V3 model. A month later, the R1 model for a more complete search made its debut. Now comes V3-0324, named after its release in March 2024.
Deep demand
Improvements bring the model to almost-parity with the GPT-4 models of Openai or Claude 2 of Openai. But, even if they are not quite the same power, they are much cheaper, according to Deepseek.
It is ultimately a huge argument of sale as the use of AI, and therefore IA costs, continue to increase. The training of AI models is notoriously expensive, and Openai and Google have huge cloud budgets that most companies could not reach without partnerships like Openai with Microsoft. This exclusivity disappears if the cheaper achievements of Deepseek become more common.
The American domination of AI models begins to slide anyway, thanks in part to Chinese startups like Deepseek. It no longer seems shocking when the hottest model emerges from Shenzhen or Hangzhou. Geopolitical considerations, as well as commercial concerns, have prompted calls to deepest at least at least the US government.
You will probably not see the latest version of Deepseek change everything for your schedule tomorrow. He suggests that the demand for a power of calculation and energy power to feed the new generation AI may not be as astounding as they are fearing.
This could also mean that the Chatbot Ai rewriting your curriculum vitae or debugging your website also speaks the Mandarin.
