DeepSeek, the Hangzhou-based AI company that rattled Silicon Valley with absurdly cheap language models, is about to get a lot less cheap. Starting August 16 at 16:00 UTC, the company will raise API prices across its V4-Flash and V4-Pro models by anywhere from 50% to more than 1,100%, depending on token type and time of day.
The increases come with a new peak and off-peak pricing structure. Using the API during busy hours will cost twice as much as using it during off-peak windows.
The numbers behind the sticker shock
For the V4-Flash model, off-peak cache-hit input tokens will cost $0.007 per million, up from $0.0028. That’s a 150% increase. Cache-miss input tokens jump from $0.14 to $0.22 per million, a roughly 57% bump. Output tokens rise from $0.28 to $0.66 per million during off-peak hours.
Peak rates double all of those off-peak figures. So output tokens during high-traffic windows will run $1.32 per million, nearly five times the previous flat rate of $0.28. The V4-Pro models follow a similar tiered structure with peak rates set at twice the off-peak level.
