DeepSeek peak-hour pricing

With the full release of the V4 series DeepSeek both updated its prices and moved to time-of-day pricing. During peak hours every token item costs twice the standard rate. The peak/off-peak split applies on weekdays only; weekends and Chinese public holidays bill at the standard rate all day. LLMTR passes this through unchanged; no margin is added to model prices.

Cutover

The new prices take effect at 2026-08-16 16:00 UTC.

• Before that instant: the current card below applies and there is no peak multiplier — every hour of the day bills the same.
• From that instant: the new card applies and the peak multiplier is in force.

From 2026-08-23 00:00 Beijing time (2026-08-22 16:00 UTC) the peak/off-peak split no longer applies on weekends at all: on Beijing Saturday and Sunday every hour of the day bills at the standard rate. The weekday schedule is unchanged.

The cutover happens on its own; which card a request bills against is decided by the moment the request reaches LLMTR.

V4.1 Flash and the retirement of the V4 Flash line

On 2026-09-10 DeepSeek released V4.1 Flash and retired the previous-generation V4 Flash line. The new model is served under deepseek/deepseek-flash, with a 1M-token context window, up to 384K output, image input, tool calls and JSON output.

The four retired identifiers remain listed on LLMTR at their own URLs, but they no longer accept requests: calling one returns an error naming its successor.

Retired identifier	Successor
deepseek/deepseek-v4-flash	deepseek/deepseek-flash
deepseek/deepseek-v4-flash-vision-exp	deepseek/deepseek-flash
deepseek/deepseek-chat	deepseek/deepseek-flash
deepseek/deepseek-reasoner	deepseek/deepseek-flash

Image input is no longer a separate experimental row: deepseek/deepseek-flash accepts images directly.

The Pro identifier

DeepSeek announced that deepseek/deepseek-v4-pro would be routed to V4.1 Flash from 2026-09-14, then withdrew that plan. The identifier continues to be served as V4 Pro and billed on the new card below. For image input or a lower unit cost, use deepseek/deepseek-flash.

The peak schedule is the same on both cards: the same two windows, the same 2x multiplier and the same weekend and public holiday exemption.

Peak windows

The windows are fixed in UTC and apply on weekdays only:

Window	UTC	Beijing (UTC+8)	Days
First	01:00 - 04:00	09:00 - 12:00	Monday - Friday
Second	06:00 - 10:00	14:00 - 18:00	Monday - Friday

Windows are half-open: 04:00 UTC already bills at the standard rate, while 03:59 UTC is still peak. Neither UTC nor Beijing observes daylight saving, so these boundaries hold year round.

The weekend is determined in DeepSeek's own civil time: Beijing Saturday and Sunday. Because both windows sit at Beijing 09:00-12:00 and 14:00-18:00 they never cross a day boundary, so the same two calendar days apply however you read the clock.

Chinese public holidays bill at the standard rate all day, even when they fall on a weekday. The dates come from the calendar China's State Council publishes each year and are read as Beijing dates. In 2026 the holidays that fall on a weekday are 1-2 January, 16-20 and 23 February, 6 April, 1, 4 and 5 May, 19 June, 25 September and 1, 2, 5, 6 and 7 October. Weekend days declared as make-up working days also bill at the standard rate.

Prices

Per 1M tokens, USD.

V4.1 Flash card

In force for deepseek/deepseek-flash from 2026-09-10.

Model	Item	Standard	Peak
deepseek/deepseek-flash	Input (cache miss)	0.15	0.3
deepseek/deepseek-flash	Cache read	0.003	0.006
deepseek/deepseek-flash	Output	0.6	1.2

New card (from 2026-08-16 16:00 UTC)

deepseek/deepseek-v4-pro bills on this card. The other rows were retired on 2026-09-10; their prices are kept here as the record of what they cost while they were live.

Model	Item	Standard	Peak
deepseek/deepseek-v4-flash	Input (cache miss)	0.22	0.44
deepseek/deepseek-v4-flash	Cache read	0.007	0.014
deepseek/deepseek-v4-flash	Output	0.66	1.32
deepseek/deepseek-v4-flash-vision-exp	Input (cache miss)	0.22	0.44
deepseek/deepseek-v4-flash-vision-exp	Cache read	0.007	0.014
deepseek/deepseek-v4-flash-vision-exp	Output	0.66	1.32
deepseek/deepseek-v4-pro	Input (cache miss)	0.66	1.32
deepseek/deepseek-v4-pro	Cache read	0.022	0.044
deepseek/deepseek-v4-pro	Output	1.98	3.96

The multiplier is uniform across every item: cache reads double during peak too.

Current card (before 2026-08-16 16:00 UTC)

This card has no peak split — the same rate applies at every hour.

Model	Input (cache miss)	Cache read	Output
deepseek/deepseek-v4-flash	0.14	0.0028	0.28
deepseek/deepseek-v4-flash-vision-exp	0.14	0.0028	0.28
deepseek/deepseek-v4-pro	0.435	0.003625	0.87

The deepseek/deepseek-chat and deepseek/deepseek-reasoner identifiers were routed to V4-Flash and billed exactly like it. All four were retired on 2026-09-10.

Which rate applies

The rate is chosen from the moment the request reaches LLMTR. A single instant is used per request, so the rate does not change mid-request. A call that starts at 09:59:59 UTC and finishes at 10:00:20 UTC is billed entirely at the peak rate rather than split across both.

Estimating cost

You can determine whether you are currently in a peak window from the UTC clock:

from datetime import datetime, timedelta, timezone

PEAK_WINDOWS = [(60, 240), (360, 600)]  # UTC minutes of day
BEIJING = timezone(timedelta(hours=8))
# Chinese public holidays that fall on a weekday (Beijing dates), 2026.
CN_HOLIDAYS = {
    "2026-01-01", "2026-01-02", "2026-02-16", "2026-02-17", "2026-02-18",
    "2026-02-19", "2026-02-20", "2026-02-23", "2026-04-06", "2026-05-01",
    "2026-05-04", "2026-05-05", "2026-06-19", "2026-09-25", "2026-10-01",
    "2026-10-02", "2026-10-05", "2026-10-06", "2026-10-07",
}

def is_peak(moment: datetime) -> bool:
    # Weekends and holidays are decided in the provider's own time zone.
    beijing = moment.astimezone(BEIJING)
    if beijing.weekday() >= 5:  # 5 = Saturday, 6 = Sunday
        return False
    if beijing.date().isoformat() in CN_HOLIDAYS:
        return False
    utc = moment.astimezone(timezone.utc)
    minute_of_day = utc.hour * 60 + utc.minute
    return any(start <= minute_of_day < end for start, end in PEAK_WINDOWS)

print(is_peak(datetime.now(timezone.utc)))

const PEAK_WINDOWS = [
  [60, 240],
  [360, 600]
];
const BEIJING_OFFSET_MINUTES = 8 * 60;
// Chinese public holidays that fall on a weekday (Beijing dates), 2026.
const CN_HOLIDAYS = new Set([
  "2026-01-01", "2026-01-02", "2026-02-16", "2026-02-17", "2026-02-18",
  "2026-02-19", "2026-02-20", "2026-02-23", "2026-04-06", "2026-05-01",
  "2026-05-04", "2026-05-05", "2026-06-19", "2026-09-25", "2026-10-01",
  "2026-10-02", "2026-10-05", "2026-10-06", "2026-10-07"
]);

function isPeak(date = new Date()) {
  // Weekends and holidays are decided in the provider's own time zone.
  const beijing = new Date(date.getTime() + BEIJING_OFFSET_MINUTES * 60000);
  const beijingDay = beijing.getUTCDay();
  if (beijingDay === 0 || beijingDay === 6) {
    return false;
  }
  if (CN_HOLIDAYS.has(beijing.toISOString().slice(0, 10))) {
    return false;
  }
  const minuteOfDay = date.getUTCHours() * 60 + date.getUTCMinutes();
  return PEAK_WINDOWS.some(([start, end]) => minuteOfDay >= start && minuteOfDay < end);
}

Both examples read UTC. Using local time gives the wrong answer depending on where your server runs.

Reducing cost

• Move batch and schedule-tolerant work outside the peak windows. 17 hours of each weekday bill at the standard rate, and weekends and Chinese public holidays bill at it entirely - 133 of the 168 hours in a week.
• Use prompt caching. The cache-read rate is far below the input rate and stays cheaper than standard input even during peak.
• Use deepseek/deepseek-flash for new integrations: it is the successor to the retired V4 Flash rows and is cheaper than they were on every item (0.15 against 0.22 on input, 0.6 against 0.66 on output).

Usage records

Every row on the Usage page stores the unit price applied to that request. Historical records are never re-priced; whatever a request was charged is what its record keeps.

Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC

On weekdays, the windows 01:00-04:00 and 06:00-10:00 UTC are peak hours, and every token item costs twice the standard rate; weekends and Chinese public holidays, determined by Beijing time, bill at the standard rate all day.