DeepSeek Launches V4 Pro With Higher Pricing
DeepSeek has officially launched its V4 Pro AI model, expanding the company’s V4 lineup with a stronger focus on agentic workloads, coding, tool use and complex reasoning. The release follows the earlier V4 Flash launch and marks a significant shift in DeepSeek’s pricing strategy.
V4 Pro targets demanding AI workloads
DeepSeek says V4 Pro is designed for advanced AI agents and demanding technical tasks. In addition, the model supports tool calls, JSON output and a 1-million-token context window. It also allows a maximum output of 384,000 tokens.
The company introduced V4 Pro alongside V4 Flash in April. However, the latest Pro release represents the general availability version of the model. Both models support thinking and non-thinking modes, while developers can access them through DeepSeek’s API.
Moreover, independent testing shows a performance gap between the two models. Artificial Analysis gave V4 Pro an Intelligence Index score of 53, compared with 40 for V4 Flash. The evaluation considers areas including coding, tool use and scientific reasoning.
Pricing rises sharply for Pro
The pricing change is a major part of the launch. DeepSeek’s previous V4 Pro rates were $0.435 per million input tokens and $0.87 per million output tokens. Meanwhile, V4 Flash cost $0.14 for input and $0.28 for output.
However, DeepSeek is introducing peak and off-peak pricing from August 16. At peak times, V4 Pro will cost $1.32 per million input tokens for cache misses and $3.96 per million output tokens. That makes the output price 14 times higher than V4 Flash’s previous $0.28 rate.
The new structure also affects cached inputs. V4 Pro’s peak cache-hit price will reach $0.044 per million tokens, while its off-peak rate will be $0.022. V4 Flash will cost $0.014 and $0.007 during peak and off-peak periods, respectively.
As a result, developers will need to consider both model performance and usage timing. Off-peak workloads could reduce costs, while high-volume agent systems may face substantially higher bills during peak periods.
DeepSeek faces a tougher AI market
The launch comes as DeepSeek faces growing competition from Chinese AI companies, including Moonshot AI, ByteDance and Alibaba. At the same time, major US AI developers continue to release increasingly capable models.
Nevertheless, DeepSeek retains a significant pricing advantage against several leading proprietary models. Its V4 strategy also combines open model weights with a large context window and compatibility with established API formats.
Therefore, V4 Pro could appeal to developers who need stronger reasoning and agent capabilities. However, V4 Flash remains the more economical choice for high-volume applications. The widening price difference means companies will increasingly have to match each model to specific workloads.
DeepSeek’s latest move also signals a broader change in the AI market. As model capabilities improve, providers are increasingly adjusting prices around inference demand, compute costs and workload intensity.
