DeepSeek V4-Pro API permanently reduced to 1/4 of the original price, cache hit only 0.025 yuan/million tokens
The price of DeepSeek V4-Pro API has been permanently reduced to 1/4 of the original price, and the cache hit is only 0.025 yuan/million tokens. The 25% discount will be permanent, and the pricing strategy is extremely aggressive.
DeepSeek V4-Pro API permanently reduced to 1/4 of the original price
DeepSeek officially announced that the DeepSeek-V4-Pro model API price will be officially adjusted to 1/4 of the original price after the 25% discount ends on May 31, 2026 - which is equivalent to making the discount permanent.
| Billing type | Original price (yuan/million Tokens) | Adjusted (yuan/million Tokens) |
|---|---|---|
| input (cache hit) | 0.1 | 0.025 |
| Input (cache miss) | 12 | 3 |
| Output | 24 | 6 |
The cache hit price is only 0.025 yuan/million Tokens, which is approximately 1/8120 of the GPT-5.5 output price ($30/million Tokens, approximately RMB 203), which is at the most aggressive level among global large model API pricing. Previously, the industry once released a signal that "high-quality models return to value pricing" - the price of some APIs of Zhipu soared by 83%, and manufacturers such as Alibaba Byte also made adjustments one after another - but DeepSeek did the opposite and made the 25% discount permanent. This move is in line with the open source strategy: to expand the developer ecosystem through the ultimate cost-effectiveness, and to build competition barriers based on the scale of call volume rather than unit price profit. For domestic competitors such as Doubao and Zhipu, the permanent price reduction of V4-Pro will directly compress the pricing space in the mid-to-high-end API market, and the end of the price war is far from coming.
Reviews