DeepSeek built a massive global name by undercutting Western tech giants with dirt-cheap AI access. However, that bargain era is officially changing. As previously teased, Hangzhou-based AI lab has officially launched its DeepSeek V4 Pro model while announcing a major pricing restructure that quadruples API costs for its flagship models during high-traffic hours.
Starting August 16, DeepSeek is introducing dynamic peak and off-peak pricing. The system aims to encourage developers to run heavy automated jobs during quieter hours. Under the new schedule, the flagship V4 Pro (specifically labeled V4-Pro-0813) will cost $1.32 per million input tokens and $3.96 for output tokens during peak hours, dropping to $1.98 for output off-peak. That is a noticeable jump from the old $0.87 promotional rate.
Meanwhile, the lighter DeepSeek V4 Flash model moves from $0.28 to $1.32 per million output tokens at peak hours ($0.66 off-peak) and $0.14 for inputs. That makes V4 Pro about 9 times pricier on inputs and 14 times higher on outputs compared to Flash.
Big benchmark jumps justify the flagship price
The higher price tag for V4 Pro comes alongside a serious performance upgrade. Back when V4 Flash launched last month, it unexpectedly outperformed an early April preview version of V4 Pro in several tests. This was a strange twist since Pro is supposed to be the premium product.
The official launch fixes that gap. According to evaluation data from Artificial Analysis, V4 Pro scored 53 on its Intelligence Index—which measures AI agent tasks, coding, and scientific reasoning—compared to a 40 score for V4 Flash. This performance leap turns V4 Pro into a much stronger choice for handling multi-step agent workflows across the company’s API, app, and web interfaces.
Managing server traffic and expansion demands
DeepSeek explained that the dynamic rate adjustments will help allocate server resources more reasonably. Despite the 4x price hike, DeepSeek remains far cheaper than most competitors. For instance, Moonshot AI’s Kimi K3 charges $15 per million output tokens, OpenAI’s flagship GPT-5.6 Sol costs $30, and Anthropic’s Fable 5 sets that price at $50. Only low-cost offerings like OpenAI’s GPT-5.6 Luna ($1.20) undercut DeepSeek V4 Flash at peak hours.
The price changes come as DeepSeek attempts to build a sustainable business by leveraging its viral buzz. DeepSeek’s R1 model launched in early 2025 went viral, but competitors like Alibaba, ByteDance, MiniMax, and Zhipu AI have made it difficult for the company to gain traction. The startup is reportedly gearing up for an IPO as soon as this year, as well as other funding rounds that have placed its valuation at $74 billion. Increasing prices helps to offset huge computing and data center expenses as the firm plans to double its engineering workforce.
The post DeepSeek’s New V4 Pro AI Model Quadruples Prices but Remains Far Cheaper Than Western AI appeared first on Android Headlines.
​Â
