Tencent Cloud's Intelligent Agent Development Platform announced that, starting at midnight on June 3, 2026, it will significantly reduce the calling price of the DeepSeek-V4 series models, bringing it in line with the official price. The cache hit price for DeepSeek-V4-Pro has been reduced by as much as 97.5%.
Specifically, the inference input and output prices for the DeepSeek-V4-Pro model have both been reduced by 75%, dropping to 0.003 yuan per thousand tokens and 0.006 yuan per thousand tokens respectively. Its cache hit price has been adjusted to 0.000025 yuan per thousand tokens; meanwhile, the cache hit price for the DeepSeek-V4-Flash model has also seen a 90% reduction, adjusted to 0.000025 yuan per thousand tokens.
As a star large model launched on April 24 this year, the DeepSeek-V4 series includes Pro and Flash versions, with a total parameter count of 1.6 trillion. It uses an advanced mixture of experts (MoE) architecture and natively supports a context length of up to one million tokens. Prior to this, DeepSeek had already changed the API price for V4-Pro from a limited-time promotion to a permanent reduction on May 22.
