DeepSeek off-peak hours and pricing in your time zone
Updated 2026-10-23. From DeepSeek's API pricing page. Local times are our conversion.
DeepSeek's off-peak hours are every hour outside 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, and the API costs half price then. Weekends and Chinese public holidays are off-peak all day. In US Eastern time, peak is 9 PM to midnight and 2 AM to 6 AM until Nov 1, then 8 PM to 11 PM and 1 AM to 5 AM, so the US workday is entirely off-peak. In India, peak runs 6:30 AM to 9:30 AM and 11:30 AM to 3:30 PM, which covers most of the morning and early afternoon.
DeepSeek peak hours in your time zone
| Time zone | Peak hours now | After clocks change |
|---|---|---|
| UTC | 01:00 to 04:00 and 06:00 to 10:00 | No change |
| China (Beijing time) | 9 AM to noon and 2 PM to 6 PM | No change |
| India (IST) | 6:30 AM to 9:30 AM and 11:30 AM to 3:30 PM | No change |
| UK | 2 AM to 5 AM and 7 AM to 11 AM | 1 AM to 4 AM and 6 AM to 10 AM |
| Central Europe (Berlin, Paris) | 3 AM to 6 AM and 8 AM to noon | 2 AM to 5 AM and 7 AM to 11 AM |
| US Eastern (ET) | 9 PM to midnight (evening before) and 2 AM to 6 AM | 8 PM to 11 PM (evening before) and 1 AM to 5 AM |
| US Pacific (PT) | 6 PM to 9 PM (evening before) and 11 PM to 3 AM (overnight) | 5 PM to 8 PM (evening before) and 10 PM to 2 AM (overnight) |
Turn on JavaScript to see whether it's peak or off-peak right now in your time zone.
Peak applies Monday to Friday by UTC date, so in the US the evening window falls on Sunday through Thursday nights. UK and European clocks go back on Oct 25, 2026 and US clocks on Nov 1, 2026. UTC hours from DeepSeek's pricing page, checked 2026-10-23; local times are our conversion. Every hour not listed is off-peak.
DeepSeek off-peak pricing per 1M tokens
| Model | Rate | Input | Cache hit | Output |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | Peak | $0.30 | $0.006 | $1.20 |
| DeepSeek V4.1 Flash | Off-peak | $0.15 | $0.003 | $0.60 |
| DeepSeek V4 Pro | Peak | $1.32 | $0.044 | $3.96 |
| DeepSeek V4 Pro | Off-peak | $0.66 | $0.022 | $1.98 |
USD per 1M tokens from DeepSeek's pricing page, checked 2026-10-23. Off-peak is exactly half of peak on every line, including cache hits.
The discount covers both models and every token type. A batch of 30M input and 6M output tokens on V4.1 Flash costs $16.20 if it runs at 9 PM ET on a Tuesday and $8.10 if it runs at 3 PM ET the same day.
Is DeepSeek off-peak the cheapest API?
Not at the very small end. Off-peak V4.1 Flash costs $8.10 for our sample month, while Claude Haiku 5.5 and GPT-6 Luna both cost $6 at list price. Where off-peak really pays is one tier up. DeepSeek V4 Pro off-peak comes to $31.68, against $120 for Claude Sonnet 5.5, Gemini 4 Argon and GPT-6.1 Sol. That's about 3.8x cheaper.
| Model | Sample month | Cache-heavy day |
|---|---|---|
| DeepSeek V4.1 Flash, peak | $16.20 | $2.50 |
| DeepSeek V4.1 Flash, off-peak | $8.10 | $1.25 |
| Claude Haiku 5.5 | $6 | $1.06 |
| GPT-6 Luna | $6 | $1.06 |
| Gemini 3.8 Flash | $45 | $7.95 |
| DeepSeek V4 Pro, peak | $63.36 | $9.94 |
| DeepSeek V4 Pro, off-peak | $31.68 | $4.97 |
| Claude Sonnet 5.5 | $120 | $19.60 |
| Gemini 4 Argon | $120 | $19.60 |
| GPT-6.1 Sol | $120 | $19.60 |
Our estimate from list prices in our data. Sample month = 30M input + 6M output tokens, no caching. Cache-heavy day = 4M fresh input + 16M cached input + 1M output, cache write and storage fees left out. Claude Haiku 5.5 uses its rate for prompts up to 100K tokens. Longer prompts cost $0.50 in, $2.50 out. Gemini 4 Argon uses its launch price through Nov 30, 2026. From Dec 1: $4 in, $20 out.
Price isn't quality, so check how the models rank before you switch. Our head-to-heads cover DeepSeek V4 Pro vs Claude Opus 5.5, DeepSeek V4 Pro vs GPT-6.1 Sol and DeepSeek V4.1 Flash vs Gemini 3.8 Flash.
How to schedule DeepSeek jobs for off-peak
- In the US: you don't need to do anything during the day. Peak sits in the late evening and the early hours, so normal working hours are already half price.
- In India, China and Europe: peak lands in your working morning. Queue big batch jobs for the afternoon (after 3:30 PM IST, 6 PM in Beijing) or the weekend.
- In cron or a job queue: schedule in UTC so clock changes can't move your window. Weekdays from 10:00 to 01:00 UTC the next day, plus 04:00 to 06:00 UTC, are off-peak, and so is all of Saturday and Sunday.
- Long requests: DeepSeek doesn't say whether a request that runs across a boundary is billed by its start or end time. Leave a few minutes of margin either side of 01:00, 04:00, 06:00 and 10:00 UTC.
There's no setting to switch on. The pricing page lists the off-peak rate by hour and doesn't mention any opt-in.
What we know and what we don't
| Question | Answer |
|---|---|
| Off-peak discount | 50% on input, cache hits and output, both models |
| Peak hours | 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday |
| Weekends and Chinese public holidays | Off-peak all day |
| Which holidays count | DeepSeek doesn't list the dates |
| Billed by request start or end time | DeepSeek doesn't say |
| Concurrency limits | 2,500 at once on V4.1 Flash, 500 on V4 Pro, per account |
| Search and Fetch tool calls | $0.75 per 1,000 requests, one price at any hour |
From DeepSeek's pricing page and rate limit docs, checked 2026-10-23.
What changes on Oct 31 for Claude Code and Codex users
If you point Claude Code, Codex or DeepSeek Harness at the DeepSeek API, the web search and fetch calls those tools make are free right now. That ends at 00:00 Beijing time on Nov 1, 2026, which is noon ET on Oct 31, per DeepSeek's Oct 17 announcement. After that each call bills at $0.75 per 1,000 successful requests.
The pricing page lists one Search and Fetch price with no peak or off-peak split, so timing your jobs won't change that part of the bill. One more housekeeping note: the old model names deepseek-v4-flash and deepseek-v4-flash-vision-exp still work, but DeepSeek now serves them with V4.1 Flash and bills the Flash price.
For the full DeepSeek price list see DeepSeek API pricing, and to weigh it against a monthly plan try the plan vs API calculator.
Sources
FAQ
When are DeepSeek off-peak hours?
Every hour outside 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays. Weekends and Chinese public holidays are off-peak all day, per DeepSeek's pricing page.
What are DeepSeek off-peak hours in EST?
Peak is 9 PM to midnight the night before and 2 AM to 6 AM ET until Nov 1, 2026, and 8 PM to 11 PM the night before and 1 AM to 5 AM ET after clocks fall back. Every other hour, including the whole US workday, is off-peak.
What are DeepSeek off-peak hours in IST?
Peak is 6:30 AM to 9:30 AM and 11:30 AM to 3:30 PM IST, Monday to Friday. Off-peak runs from 3:30 PM to 6:30 AM the next morning, plus 9:30 to 11:30 AM, and all weekend.
How much is the DeepSeek off-peak discount?
50%. V4.1 Flash drops from $0.30 to $0.15 input and $1.20 to $0.60 output per 1M tokens, and V4 Pro from $1.32 to $0.66 input and $3.96 to $1.98 output.
Are DeepSeek API weekend rates cheaper?
Yes. Saturday and Sunday are off-peak for the full day in UTC, so the whole weekend bills at half price.
Do I need to turn on DeepSeek off-peak pricing?
No setting is mentioned. DeepSeek's pricing page sets the rate by the hour, so requests in off-peak hours get the lower price.
Related
- DeepSeek API pricing
- DeepSeek vs Claude API cost
- Cheapest LLM API
- DeepSeek V4 Pro vs Claude Opus 5.5
- DeepSeek V4.1 Flash vs Gemini 3.8 Flash
- Qwen3.8 Max vs DeepSeek V4 Pro
- All AI API prices
- Plan vs API calculator
- Best model for coding
More guides
- OpenAI vs Gemini API pricing
- Perplexity limits explained
- Recursive Flint: can a free local model replace a $20 coding plan?
- Windsurf (Devin) limits explained
- Windsurf usage limit reached: what to do
- What $100 a month AI plans include
Not sure a plan beats the API for you? Try the plan vs API cost calculator.