Costs at off-peak hours shall be half the peak-hour pricing.
DeepSeek made its title providing far cheaper AI companies than its pricier western opponents, however the low cost experience seems to be at an finish. The corporate has began telling clients to arrange for vital value rises with the appearance of its newest mannequin. With the announcement of DeepSeek V4 Professional, the corporate is elevating its API pricing fourfold. The corporate stated it is adopting a brand new peak and off-peak pricing to “allocate sources extra fairly.”
Beginning on August 16, the DeepSeek V4 Professional mannequin will value $3.96 for 1 million output tokens at peak hours, greater than 4 occasions the present fee of $0.87. It’s going to value you half that quantity ($1.98) throughout off-peak hours. In the meantime, the DeepSeek V4 Flash mannequin will set you again $1.32 for 1 million output tokens (from $0.28) throughout peak hours and $0.66 throughout off-peak.
To be truthful, the present API pricing was imagined to be only a promotion ending on Could 31. DeepSeek introduced that month that it was going to make the discounted costs everlasting, however it finally determined to proceed with the worth hikes.
Regardless that DeepSeek will quickly value 4 occasions extra, its V4 fashions are nonetheless much more inexpensive than a lot of its opponents. Kimi K3, a mannequin by Chinese language firm Moonshot, prices $15 per 1 million output tokens. OpenAI’s most superior mannequin, GPT-5.6 Sol, prices $30 for 1 million output tokens. Nonetheless, its low value providing, GPT-5.6 Luna, prices $1.20, lower than the DeepSeek V4 Flash at peak hours.
