DeepSeek launches V4 Pro, a premium version of its large language model, with prices up to 14 times higher than the V4 Flash, seeking to establish a new hierarchy in its AI product suite amid fierce competition and performance gains.
DeepSeek has unveiled V4 Pro, a higher-end version of its large language model, and is charging a markedly steeper rate than for V4 Flash as it tries to convert technical progress into a premium product. According to Artificial Analysis, V4-Pro-0813 costs $1.32 per million input tokens and $3.96 per million output tokens, compared with $0.14 and $0.28 for V4 Flash, putting the flagship model at about nine times the input price and 14 times the output price. DeepSeek said the model is available through its API, app and web products and that it improves performance for AI agents, a category of systems designed to plan, use tools and complete multi-step tasks.
Independent testing suggests the launch is more than a branding exercise. Artificial Analysis said the reasoning version of V4 Pro scored 53 on its Intelligence Index, well above V4 Flash at 40. The benchmark combines assessments of agentic work, tool use, coding, scientific reasoning and long-context performance. The pricing and score gap also help explain why the company is now presenting V4 Pro as the more capable commercial tier after V4 Flash unexpectedly outperformed an April preview of the model in several third-party tests.
That earlier result had been unusual because the Pro line was meant to be DeepSeek’s strongest offering. The latest release appears to narrow that gap and restore the intended hierarchy between the two products. DeepSeek also said it will increase API pricing for both V4 Pro and V4 Flash and add peak and off-peak rates, a sign that it is moving towards a more complex commercial model as demand and usage patterns evolve.
The launch comes as DeepSeek tries to keep pace in a fiercer domestic market. The company rose to prominence after its R1 model drew global attention in early 2025, fuelling debate over whether top-tier AI could be built at far lower cost than in the US. Since then, rivals including Moonshot AI, Zhipu AI, MiniMax, Alibaba and ByteDance have intensified competition. Reuters reported in July that DeepSeek was preparing a new fundraising round at a valuation of about $74 billion after raising about $7.4 billion in its first outside financing in June, a shift for a company that had long avoided external capital. Reuters also reported that DeepSeek aims to at least double staff across several teams and has been hiring chip-design engineers as it seeks greater independence from suppliers such as Nvidia and Huawei.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
