When DeepSeek released the DeepSeek V4 Pro model before, many developers complained because the price was too high. At that time, DeepSeek also explained that the high price was due to insufficient computing power. When the computing power is sufficient, the price will be adjusted so that everyone can use it with confidence. To comfort developers at that time, DeepSeek also launched a limited-time promotion.
The limited-time promotion is provided at a 25% discount on the model price and is valid until May 31, 2026. The end of the limited-time promotion is very close. Unexpectedly, Deepin Search announced at this time that after the promotion ends, it will adjust the model price to 1/4 of the original price, which means that the promotional price will become the regular price in the future, and this price is relatively cheap.
According to the current price list, the DeepSeek V4 Pro million input hit cache is only 0.025 yuan, the million input miss cache is only 3 yuan (original price 12 yuan), and the million output is only 6 yuan (original price 24 yuan). For most developers, this price can be used with confidence, so it can be configured into various programming development tools and used smoothly.
DeepSeek V4 series models support up to 1M length and 384K output. The Flash model has a concurrency limit of 2500 per account, and the Pro model has a concurrency limit of 500. For corporate and team customers, if the concurrency limit is not enough, you can contact DeepSeek to apply for a higher concurrency limit. You can also configure different user functions under the same API through the new USER ID function.
DeepSeek API platform: https://platform.deepseek.com/usage
DeepSeek API pricing: https://api-docs.deepseek.com/zh-cn/quick_start/pricing/
DeepSeek API concurrency: https://api-docs.deepseek.com/zh-cn/quick_start/rate_limit