DeepSeek is set to release a new AI model to rival advances in OpenAI’s ChatGTP and Google’s Gemini.

The newer model will follow its first AI large language model that came out in January 2025, striking a major challenge to OpenAI due to its reduced compute and energy requirements.

The launch of the firm’s first model was deemed a “wake-up call” for the wider and US-dominant tech industry by US President Donald Trump.

The new V4 model from DeepSeek is set to be multimodal, with the ability to generate text, pictures, and video, according to the Financial Times.

However, the Chinese startup has circumvented the status quo, and yet not shown its latest model to any US chipmakers, according to Reuters, but has shared the tech with Huawei and other local chip suppliers.

Models are typically shared with hardware firms to ensure that the software is compatible, but DeepSeek has reportedly not done so with US-based chipmakers, in a move that could be targeting the US’s dominance in the industry.

DeepSeek’s R1 model rocked the AI world; it was not only the first Chinese competitor that could rival other AI giants, but was created at a fraction of the cost of ChatGPT in terms of expense and compute power, was open-source, and free to use.

Recommended reading

• Oracle Shares Drop 10% – Is the AI Bubble Bursting?
• Anthropic Reaches $380bn as AI Investment Frenzy Rolls on
• Comment | What To Expect if the AI Bubble Bursts
• Nvidia Shatters Forecasts as AI Bubble Debate… Bubbles

The models’ release sent shockwaves through the AI ecosystem, causing the US-dominant AI market to plummet. Nvidia’s stock share price dropped by 20%, while major players such as Oracle, Broadcom, Microsoft, Alphabet, and Amazon also saw their share prices drop.

While DeepSeek has released other models since, the V4 model will be the first since R1 that will not only be available on Hugging Face and GitHub.