DeepSeek has officially launched its highly-anticipated artificial intelligence (AI), DeepSeek V4, offering a range of advanced capabilities.
DeepSeek V4, a preview variant, is currently accessible to use, reportedly better optimised for the chips that China is creating domestically.
According to DeepSeek, the recently introduced model “features an ultra-long context of one million words, achieving leadership in both domestic and open-source fields across agent capabilities, world knowledge, and reasoning performance".
DeepSeek V4 is currently accessible as DeepSeek V4-Pro and DeepSeek V4-Flash.
DeepSeek V4-Pro comes with a “maximum reasoning effort mode”, which is claimed to “significantly advance the knowledge capabilities of open-source models, firmly establishing itself as the best open-source model available today”.
The launch marked the first time a Chinese rival has competed with the most cutting-edge AI-powered model from other US giants.
The latest AI model launch comes amid increasing semiconductor export restrictions from the US to China, particularly high-end graphical processing units (GPUs), which lays the foundation of AI models.
Though the company didn’t reveal which chip system it used to train the V4 models; however, it said that its software components are particularly designed to work with both Huawei and Nvidia chips.
Currently, DeepSeek has yet to release more details of the recently launched version.
DeepSeek further claimed the new version achieves a “dramatic leap in computational efficiency” with its ability to process and understand the context of up to 1 million tokens.
In terms of understanding the context of long strings of text, DeepSeek V4-Pro outshines Google’s Gemini-3.1-Pro, the company says.
