DeepSeek releases official V4 Pro model as it steps up expansion

Chinese AI startup DeepSeek formally unveiled its V4 Pro model on Thursday, aiming to regain market share amidst intense competition from domestic rivals. The V4-Pro-0813 version boasts significantly improved agent capabilities and is accessible via API, APP, and Web platforms. DeepSeek announced plans to raise API prices for its V4 Pro and V4 Flash models, introducing peak and off-peak pricing tiers.

The release comes in the wake of DeepSeek's V4 Flash model unexpectedly outperforming the April preview of V4 Pro in several independent tests, despite being designed as the less powerful model. This unexpected success suggests rapid advancements in DeepSeek's technology between the preview and official launch. DeepSeek gained prominence after its R1 model went viral in early 2025, sparking debates about the feasibility of building advanced AI systems at lower costs compared to U.S. counterparts.

Since then, DeepSeek has faced competition from numerous Chinese AI firms, including Moonshot AI, Zhipu AI, MiniMax, Alibaba, and ByteDance. The company has also struggled to translate its prominence into sustained business growth. Reuters reported in July that DeepSeek was preparing to raise a new round of funding at a valuation of approximately $74 billion, following a $7.4 billion investment round in June.

This marks a shift for a company that had previously avoided external capital, reflecting the rising costs of competing in the AI industry, which necessitates substantial investments in computing hardware, data centers, and specialized personnel. DeepSeek has stated its goal to at least double its workforce in various departments, including data-center and AI-agent teams.

The company is also reportedly increasing its private hiring of chip-design engineers to develop its own AI chip, potentially reducing dependence on suppliers like Nvidia and Huawei.