DeepSeek had introduced time-based pricing earlier this month. At midnight Beijing time on Aug. 23, the weekend pricing change goes live; any charges made before that point are still covered by the old price rules.
This move is a continuation of DeepSeek's effort to balance compute with usage demand. The company's time-based pricing model, which sets peak rates at double the off-peak rate, was created to encourage developers to schedule heavy workloads during lower-traffic periods. For instance, the DeepSeek-V4-Pro model costs about 27 yuan (roughly $4) per million tokens for each call made during peak hours.
The peak pricing periods are 09:00-12:00 and 14:00-18:00 Beijing time. During these hours, the price is twice the off-peak rate. That means the off-peak price for that model would be 13.5 yuan per million tokens, since peak is double the off-peak. With the weekend change, developers can avoid the 2× multiplier entirely on Saturdays and Sundays.
For teams working across time zones, the weekend adjustment is especially beneficial. Because Beijing time is ahead of Western regions, a weekend in Beijing often covers parts of Friday and Saturday in those regions, so the effective off-peak period expands beyond just two calendar days. This simplifies planning: developers no longer need to watch the clock to avoid peak charges on weekends.
Since weekend rates are dropping, grab the free Always Be Buying E-Book to build wealth steadily.
The company had set peak windows to manage server load during high-demand hours. Now, by treating weekends as off-peak, DeepSeek aligns its pricing with typical usage patterns, as many developers reduce their API calls on Saturdays and Sundays. This could also encourage more experimentation and testing during low-traffic periods, which may improve overall service reliability.
Any usage before midnight Beijing time on Aug. 23, regardless of the user's local time, will still be billed under the previous rate schedule. So developers should check their usage logs and billing timestamps to avoid surprises. After that moment, all weekend usage will be charged at the lower off-peak rate.
By shifting batch jobs to off-peak windows, developers can cut costs significantly while easing server congestion. The company's move also signals that it expects weekend traffic to remain low, so it's willing to forgo higher margins to encourage more sustained usage.
This kind of flexible pricing is becoming common among AI infrastructure providers, and DeepSeek's approach may set a precedent for other model makers. For teams that rely on DeepSeek's API for prototyping or testing, the weekend change means they can experiment freely without worrying about surcharges. And for production workloads, the savings could add up quickly if they can be scheduled over the weekend.
The new rate applies to all models, but the impact is most visible on the high-end V4-Pro. Since the weekend is now off-peak, any usage on Saturday or Sunday - even during the usual 09:00-12:00 and 14:00-18:00 Beijing windows - will be billed at the base rate.
That's a 50% discount compared to weekday peak hours. Developers in Western regions might especially benefit, as Beijing's weekend overlaps with their Friday evenings and Saturday mornings, giving them a wider window to run heavy tasks.
Any usage before midnight Beijing time on Aug. 23 is still subject to the old rates, so plan accordingly. After that, the weekend off-peak pricing is automatic - no need to adjust code or request a discount. Just make sure your calls are timestamped after the cutoff.
The weekend pricing is a simple but effective way to align costs with real-world demand. DeepSeek's decision to lower rates during low-traffic periods is a smart move for both the company and its users, and it will be interesting to see if other providers follow suit.
With off-peak pricing now on weekends, get the free Always Be Buying E-Book for consistent investing.