DeepSeek has unveiled the preview version of its new model series, DeepSeek-V4, making it available to users and developers while also open-sourcing the system, the company announced on Thursday.

The release marks a significant upgrade in the firm’s large language model lineup, with DeepSeek-V4 offering an ultra-long context window of up to one million tokens and improved performance across agent capabilities, world knowledge and reasoning. The company claimed the model delivers leading results within China and the broader open-source AI ecosystem.

Announcing the launch in a post on X, DeepSeek said:
“DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.”

Also read | DeepSeek releases open-source V3.2 models rivaling GPT-5, Gemini

The company further highlighted the capabilities of its flagship variant, DeepSeek-V4-Pro, stating that it offers enhanced agentic capabilities and achieves state-of-the-art performance in open-source agentic coding benchmarks. It also emphasised the model’s rich world knowledge, noting that it leads all current open models and trails only Gemini-3.1-Pro. In addition, the company said DeepSeek-V4-Pro delivers world-class reasoning, outperforming existing open models in mathematics, STEM and coding tasks, while rivalling top closed-source systems.

Also read | Meet Chinese prodigy Luo Fuli: The brain behind Deepseek's success

Users can access the new models through the official website or app, while developers can integrate them via updated APIs by specifying model names. DeepSeek said both versions support a context length of one million tokens and a maximum output of up to 384,000 tokens. The models also support JSON output, tool calls and agent-style interactions, reflecting a growing industry shift toward autonomous AI systems capable of executing complex tasks.

In addition, the company has optimised DeepSeek-V4 for compatibility with mainstream agent tools, including Claude Code, OpenClaw, OpenCode and CodeBuddy, improving performance in coding and document-generation tasks.

Founded in 2023 by Liang Wenfeng and backed by hedge fund High-Flyer, DeepSeek has rapidly emerged as a key player in the global AI race. Its earlier model, DeepSeek-R1, delivered performance comparable to leading systems such as GPT-4, underscoring the company’s ambitions to compete with top-tier closed-source AI platforms.

At home, DeepSeek's AI tools have been widely adopted by Chinese municipalities and healthcare institutions as well as the financial sector and other businesses.

This has been partly driven by DeepSeek's decision to make its systems open source, with their inner workings public -- in contrast to the proprietary models sold by OpenAI and other Western rivals.

But the White House has accused Chinese firms of vying to "steal" American technology, ahead of an expected summit between Donald Trump and Xi Jinping in Beijing next month.

"The US has evidence that foreign entities, primarily in China, are running industrial-scale distillation campaigns to steal American AI," Trump's science and technology chief advisor Michael Kratsios said in a post on X.

Distillation is a common practice within AI development, often used by companies to create cheaper, smaller versions of their own models.

DeepSeek's Friday announcement also came as Meta said it planned to cut a tenth of its staff as it looks for productivity gains from the rest of the workforce while investing heavily in artificial intelligence. Reports said Microsoft was also looking to trim its ranks.