DeepSeek releases long-awaited new AI model V4, promising to rival established players in the industry.

DeepSeek has officially unveiled its V4 AI model, which is touted as a major challenger to established players like OpenAI and Google. The model features a groundbreaking one million token context length, allowing it to handle extensive text inputs, including entire documents and codebases.
The company claims that V4 incorporates a novel approach called Hybrid Attention Architecture, enhancing its ability to maintain conversational context and improve reasoning. This follows the release of its previous model that shook the AI landscape with impressive performance at a lower price point.
“DeepSeek has the best agentic coding capability among open-source models,” the company stated, emphasizing upgrades in the model’s reasoning abilities and autonomy for tasks like coding. The V4 utilizes domestic chips from Huawei and Cambricon, contrasting sharply with its predecessor, which relied on Nvidia hardware.
In a strategic partnership, Huawei aids DeepSeek’s computational requirements through its ‘Supernode’ technology, showcasing the collaboration between leading Chinese tech entities. However, analysts have expressed that the impact of V4 on the market may not replicate the frenzy caused by its predecessor.
With the AI race intensifying globally, particularly between the US and China, DeepSeek's innovative strides signify a pivotal moment in open-source AI development, offering high-caliber capabilities to the wider audience.
“The AI race has intensified the rivalry between China and the United States,” underscoring the competitive tensions surrounding technological advancements in the field.

DeepSeek has launched its highly anticipated AI model V4, proposing significant cost reductions and a one million token context length, rivaling industry leaders like OpenAI and Google. This comes after a year of innovation following the company's previous low-cost model that matched US competitors.