Chinese AI startup DeepSeek formally released its V4 Pro model on Thursday, pricing it several times higher than its V4 Flash model as the company looks to convert stronger benchmark results into a premium flagship product.
According to independent benchmarking firm Artificial Analysis, DeepSeek’s V4-Pro-0813 is priced at $1.32 per million input tokens and $3.96 per million output tokens — roughly 9 times the input cost and 14 times the output cost of V4 Flash, which Artificial Analysis lists at $0.14 and $0.28 respectively.
DeepSeek said V4-Pro-0813 brings substantial improvements to its capabilities for AI agents, and the model is now available through the company’s API, app, and web products. DeepSeek also said it will raise API pricing across both V4 Pro and V4 Flash, while introducing peak and off-peak pricing tiers.
A Premium Built on an Unusual Backstory
The steep pricing reflects DeepSeek placing a larger premium on its top-tier model following an unusual episode last month: V4 Flash, released as the lower-cost option, unexpectedly outperformed an April preview version of V4 Pro in several independent tests. That result stood out because Pro is meant to be the company’s more capable offering, suggesting DeepSeek made rapid improvements to its technology between the April preview and this week’s official launch.
Independent test results now suggest the official V4 Pro release represents a meaningful step up from that earlier preview. Artificial Analysis scored the reasoning version of V4 Pro at 53 on its Intelligence Index, compared to 40 for V4 Flash — an index built from evaluations across nine areas, including agentic task performance, tool use, coding, scientific reasoning, and long-context handling.
DeepSeek Looks to Regain Ground in a Crowded Field
DeepSeek became one of China’s most closely watched AI companies after its R1 model went viral in early 2025, sparking broader questions about whether powerful AI systems could be built at a fraction of the cost associated with US competitors.
Since then, its early lead has faced growing pressure from a wave of releases by Chinese rivals, including Moonshot AI, Zhipu AI, MiniMax, Alibaba, and ByteDance.
The company has also had to grapple with turning its early prominence into a durable business. Reuters reported in July that DeepSeek was planning a new funding round at a valuation of roughly $74 billion, just weeks after raising about $7.4 billion in its first outside financing round in June — a notable shift for a company that had long avoided external capital.
The fundraising push reflects the rising cost of competing in AI, an industry that demands heavy investment in computing chips, data centers, and specialized talent. DeepSeek has said it aims to at least double staffing across several departments, including its data-center and AI-agent teams.
Reuters also reported in July that the company had ramped up private hiring of chip-design engineers to develop its own AI chip — an effort that could reduce its dependence on suppliers such as Nvidia and Huawei.
