What Is DeepSeek Off-Peak Pricing?
DeepSeek off-peak pricing bills API usage at 50% of the peak rate during every hour outside two Beijing-time windows: 09:00–12:00 and 14:00–18:00.
The exact hours
DeepSeek splits the day in half. Peak hours are Beijing time 09:00–12:00 and 14:00–18:00 and bill at full price; everything else is billed at 50% off. Note the windows are Beijing time — if your servers run in UTC, the discount hours land at 01:00–04:00 UTC and 06:00–10:00 UTC.
There are no weekend or holiday exceptions in the published model — the same two windows apply every day.
Which models and prices it applies to
Off-peak pricing applies to DeepSeek's own models, which carry a peak/off-peak price pair: V4 Flash (¥2.00 → ¥1.00 input, ¥8.00 → ¥4.00 output at peak/off-peak) and V4 Pro (¥9.00 → ¥4.50 input, ¥27.00 → ¥13.50 output). Cache-hit input is likewise halved (¥0.04 → ¥0.02 for Flash).
Competitor models billed through other providers follow their own pricing; DeepSeek's off-peak discount is specific to DeepSeek-owned models.
How to use it
Move batch jobs, cron workloads, and nightly summarization pipelines to off-peak hours. A nightly pipeline generating 100M output tokens on V4 Flash costs ¥800 at peak — or ¥400 after 8 PM. Same tokens, same model, zero code changes, one scheduling tweak.
Interactive traffic naturally lands in peak windows; if you can't shift the traffic itself, consider caching to offset the input-side peak cost. Off-peak scheduling and a high cache hit rate are the two biggest levers on a DeepSeek bill.
Frequently Asked Questions
What are DeepSeek's exact peak hours?
Beijing time 09:00–12:00 and 14:00–18:00 every day. All other hours are billed at 50% off the peak price.
Does off-peak pricing apply to all DeepSeek models?
Yes — DeepSeek's own models (V4 Flash, V4 Pro) carry peak/off-peak price pairs on input, output, and cache-hit input alike. Competitor models billed through other providers follow their own pricing.
How much can I save by running at off-peak?
50% off every billable rate. A nightly 100M-output-token V4 Flash job costs ¥800 at peak but ¥400 off-peak — the same tokens at half the price.
Dive deeper
Related terms
Recommended Tools We ARE USING
We are using these tools ourselves for Development / Deployment. Check out for more details.
Opencode Go
You will receive $5 when subscribed, which directly offsets the 1st month's Go fee.
Could be the cheapest option for calling DeepSeek V4 Flash model in the world.
Vultr
RReceive $300 to test out VVultr platform
Referred user must be active 30+ days and spend $10–$25
Railway
Deploy anything without the complexity
Connect your repo, Railway handles the rest.
腾讯云 Tencent Cloud
Cloud servers, cloud databases, COS, CDN, SMS and other cloud products are on special offer now.
AWS high-quality alternative, we use COS as a replacement for S3
硅基流动 SiliconFlow
¥16 universal voucher for all platforms, valid for 180 days from the date of receipt
A comprehensive product matrix supporting the full-process implementation of AI applications.
Warp
One of the best AI terminal tools in the world
Interestingly, we can directly use various top-tier models in Warp without leaving the terminal.