Chinese startup DeepSeek has released a new artificial intelligence model with “drastically reduced” costs, more than a year after it shocked the tech world with a low-cost inference model that matched the capabilities of American rivals.
The race to develop artificial intelligence has further deepened the rivalry between China and the United States, with the White House on Thursday accusing Chinese entities of an extensive attempt to steal AI technology.
The Hangzhou-based company became globally famous in January last year when it introduced a chatbot based on the R1 model, which challenged the US dominance in this strategic sector.
The new model, DeepSeek V4, according to the company’s announcement, has “ultra-long context of one million words” and offers “world-leading performance with drastically reduced computational and memory costs.”
The model comes in two versions – V4-Pro and V4-Flash – where the Flash version is more efficient and cheaper, while the Pro version has a significantly larger number of parameters and more advanced inference capabilities.
Experts have assessed that this model could represent a “turning point” in the industry, especially due to the reduction in the cost of processing long texts and wider application in commercial and research areas.
– This solves the long-standing problems of slower processing and high costs for long contexts – said one technology analyst, adding that the technology could become an industry standard.
The new model is also optimized for popular AI tools and systems, and in some tests it reportedly reaches or approaches the results of leading closed models such as Google’s systems.
Last year’s “DeepSeek shock” caused shares in the AI sector to fall, as it showed that China could compete with the US with significantly fewer resources.
However, the company’s success has also raised questions about data privacy and censorship, as the chatbot sometimes avoids responding to politically sensitive topics, including events like the Tiananmen protests.
At the same time, the US administration claims that Chinese companies are conducting “industrial campaigns” to steal AI technology, further straining relations ahead of possible political meetings between the leaders of the two countries.
(Vijesti.ba)
