DeepSeek now bills Chinese public holidays at off-peak rates
- What happened
- DeepSeek's API rate card now bills Chinese public holidays entirely at off-peak rates, including weekday hours that would otherwise be peak.
- Why it matters
- On those weekdays, work run 01:00-04:00 or 06:00-10:00 UTC costs half what it does on a normal weekday. Per-token rates did not change.
- What to do
- If you schedule DeepSeek batch work around the peak window, treat Chinese public holidays as off-peak days.
Our take: a small, real saving for anyone who batches DeepSeek work, and nothing more. Chinese public holidays now bill at off-peak rates around the clock, which halves the cost of weekday work in the 01:00-04:00 and 06:00-10:00 UTC windows on those days. Per-token prices did not move, so our verdicts on DeepSeek V4.1 Flash (conditional), DeepSeek V4 Pro (recommended) and DeepSeek V4 Flash (recommended) stand.
What happened
DeepSeek's API rate card now takes Chinese public holidays out of its peak window. The footnote reads: "Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other hours are off-peak, including weekends and Chinese public holidays in full."
When we read the same footnote on 9 September, it said only "Monday through Friday (all other hours are off-peak)". DeepSeek does not date the change, and the page lists no holiday calendar.
Why it matters
Off-peak is half the peak rate, and the rates themselves are unchanged:
| Model id | Serves | Input, off-peak / peak (per 1M) | Output, off-peak / peak (per 1M) |
|---|---|---|---|
| deepseek-flash | DeepSeek-V4.1-Flash | $0.15 / $0.30 | $0.60 / $1.20 |
| deepseek-v4-pro | DeepSeek-V4-Pro-0813 (per the rate card) | $0.66 / $1.32 | $1.98 / $3.96 |
On a Chinese public holiday that falls on a weekday, a peak-hour V4 Pro job now costs $1.98 per 1M output tokens instead of $3.96. The legacy deepseek-v4-flash id is still accepted, but the rate card says its requests are served by DeepSeek-V4.1-Flash and billed at the Flash price.
V4 Pro: three DeepSeek pages, three answers
DeepSeek's own pages still do not agree on what the deepseek-v4-pro id serves after 14 September:
- Launch post (10 September): "Starting at 04:00 UTC on Sept 14, 2026, all deepseek-v4-pro requests will route to V4.1-Flash at V4.1-Flash rates." It adds that this "will continue until V4.1-Pro launches."
- API changelog (entry dated 10 September): DeepSeek will "continue providing API services for DeepSeek V4 Pro after September 14, 2026, with the billing method remaining unchanged."
- Rate card (read 23 September): still lists deepseek-v4-pro as DeepSeek-V4-Pro-0813 at the prices above, but no longer carries the sentence about continuing the service.
We are not picking a side the sources do not support. Our V4 Pro entry keeps its rates and says nothing about routing.
What changes for you
- If a scheduler shifts DeepSeek work out of the weekday peak window, mark Chinese public holidays as off-peak days.
- Check holiday dates against a calendar you trust: DeepSeek names the rule but lists no dates.
- If you depend on V4 Pro behaviour specifically, confirm which model answers your own requests until DeepSeek publishes one answer.
What to do
- 1 Add Chinese public holidays to the off-peak days in any scheduler that shifts DeepSeek work out of 01:00-04:00 and 06:00-10:00 UTC on weekdays.
- 2 Check the holiday dates against a calendar you trust: DeepSeek's rate card names the rule but lists no dates.
- 3 If you depend on DeepSeek V4 Pro specifically, confirm which model answers your requests. DeepSeek's own pages still disagree on whether deepseek-v4-pro reroutes to V4.1-Flash.
Affected tools & models
Never need to catch up again
The weekly delta: only verdict changes and act-now items. No digest filler.