The article primarily introduces DeepSeek's recent major announcement: permanently lowering the API price for the DeepSeek-V4-Pro model.
Author and source: Li Yuxia
On May 22, DeepSeek officially announced a major update: the API pricing for the DeepSeek-V4-Pro model will be permanently significantly reduced, turning what was previously a limited-time offer into a permanent price cut.
The new specific rates will officially take effect on May 31, 2026, with the adjusted price set at one-quarter of the original price (i.e., a 25% discount), establishing a new global low for large model pricing.
What does this price reduction mean?
For regular users:
If you normally only chat using DeepSeek’s official website or app, it remains completely free and is not affected by this API price adjustment.
For developers and enterprises:
This is a substantial, tangible benefit. Especially for those who need to heavily utilize APIs for coding, building agents, or handling high cache-hit-rate scenarios (such as RAG knowledge bases), costs have dropped by more than 90%. Some netizens have even exclaimed that they’ve achieved “Token freedom.”
Industry comparison:
On the international market, this rate is highly competitive. For example, OpenAI’s recently released GPT-5.5 has an output pricing of approximately $30 per million tokens (about RMB 216), which is more than 30 times more expensive than DeepSeek.
Why is DeepSeek daring to lower prices against the market trend?
At a time when many major model providers in the industry are raising prices or introducing fees, DeepSeek dares to slash its pricing to “rock-bottom” levels, primarily thanks to two key advantages:
Technical breakthrough reduces costs:
The next-generation V4-Pro model reduces computational requirements per inference to just 27% of the previous generation and cuts memory usage to one-tenth, thanks to architectural innovations such as a hybrid attention mechanism. This significant cost reduction makes the permanent price cut commercially viable.
Backed by substantial funding:
At the same time as announcing price reductions, DeepSeek is advancing a funding round of up to RMB 70 billion, with founder Liang Wenhong even planning to personally invest approximately RMB 20 billion. This indicates that the company’s current strategic priority is to expand technological boundaries and secure a leading position in AGI (Artificial General Intelligence), rather than focusing on short-term monetization.
In short, DeepSeek is using ultra-low pricing to attract developers worldwide and accelerate the building of its own AI ecosystem. If you or someone you know needs to integrate a large model API, now is an excellent time to get started!
Data and image sources: DeepSeek official website, publicly available online information, etc.