To further lower the innovation threshold for enterprises and developers, the Tencent Cloud Intelligent Agent Development Platform officially announced a major pricing adjustment on Tuesday. Starting at 0:00 Beijing Time on June 3, the platform will significantly reduce the prices of the entire V4 series of large models under DeepSeek.

image.png

Core model prices drop dramatically

In this price adjustment, the main model DeepSeek-V4-Pro has seen a huge reduction of 75% in both the input and output costs for inference, significantly reducing users' daily call costs. More notably, the cache hit cost of this model has dropped by as much as 97.5%, offering high cost-effectiveness for high-frequency reuse scenarios.

Additionally, the lightweight and agile DeepSeek-V4-Flash model has also received corresponding discount adjustments. The cache hit cost of this model has been reduced by 90%, allowing developers to enjoy more competitive computing costs when building low-latency, high-concurrency intelligent applications.

Seamless integration with the latest technological achievements

As the technical star of this price cut, the DeepSeek-V4 series of large models attracted industry attention from the moment of its release, thanks to its mixture-of-experts (MoE) architecture. This series of models has a total parameter count of 1.6 trillion and natively supports ultra-long context processing capability of up to 1 million tokens.

The recent price reduction by the Tencent Cloud Intelligent Agent Development Platform not only complements its strategy of ending the free public test and moving towards formal commercialization, but also directly brings the call cost level to that of DeepSeek's permanent price reduction policy. This move is bound to further intensify competition in the domestic cloud computing power market during the stage of large model application deployment.