DeepSeek unveils V4 preview model with Huawei chip support

By Estefano Gomez · Published April 24, 2026 · 2 min read · Source: Crypto Briefing

DeepSeek unveils V4 preview model with Huawei chip support

The open source model ships with 1 million token context and Huawei chip support, marking DeepSeek’s biggest release since R1.

Add us on Google by Estefano Gomez Apr. 24, 2026

DeepSeek launched the preview version of its V4 AI model on Friday, introducing a new open source system with a 1 million token context window and support for Huawei’s domestic AI chips.

The Chinese AI startup released the model in two versions. DeepSeek V4 Pro has 1.6 trillion total parameters and 49 billion active parameters, while DeepSeek V4 Flash has 284 billion total parameters and 13 billion active parameters, according to DeepSeek’s official release.

DeepSeek said V4 Pro is designed for stronger performance on complex tasks, while V4 Flash is positioned as a faster and cheaper option. The Pro version performs well in areas such as agentic coding and competitive programming, though it still trails top closed source models from companies such as Google and OpenAI.

The launch also marks DeepSeek’s first major model adapted for Huawei hardware. Huawei’s Ascend supernode, powered by Ascend 950 AI chips, will fully support DeepSeek V4 following the preview release, underscoring China’s effort to reduce reliance on foreign AI infrastructure.

That hardware shift is central to the release. DeepSeek’s earlier models were associated with Nvidia chips, but US export restrictions have pushed Chinese AI developers to rely more heavily on domestic alternatives. Huawei’s infrastructure also contributed to training the Flash version of V4.

The release comes more than a year after DeepSeek’s R1 model jolted global AI markets by showing that a Chinese lab could build competitive models at lower reported costs. V4 is being framed as another challenge to US AI companies, with DeepSeek claiming performance against top tier closed source rivals while keeping the public version open source.

DeepSeek has made V4 available through chat.deepseek.com under Expert Mode and Instant Mode, while its API was also updated on Friday.

Disclosure: This article was edited by Estefano Gomez. For more information on how we create and review content, see our Editorial Policy.

This article was originally published on Crypto Briefing and is republished here under RSS syndication for informational purposes. All rights and intellectual property remain with the original author. If you are the author and wish to have this article removed, please contact us at [email protected].