DeepSeek API pricing: peak and off-peak prices per 1M tokens
Updated · 3 min read
DeepSeek bills per million tokens and charges half price outside peak hours. On 2026-11-18, deepseek-flash cost $0.30 per million input tokens and $1.20 per million output tokens at peak, and $0.15 / $0.60 off-peak. The larger deepseek-v4-pro cost $1.32 / $3.96 at peak and $0.66 / $1.98 off-peak. Input that DeepSeek reads from its cache costs a small fraction of that.
DeepSeek model prices
These are the prices per 1M tokens on DeepSeek's pricing page on 2026-11-18:
| deepseek-flash, peak | deepseek-flash, off-peak | deepseek-v4-pro, peak | deepseek-v4-pro, off-peak | |
|---|---|---|---|---|
| Input, cache hit | $0.006 | $0.003 | $0.044 | $0.022 |
| Input, cache miss | $0.30 | $0.15 | $1.32 | $0.66 |
| Output | $1.20 | $0.60 | $3.96 | $1.98 |
- deepseek-flash runs DeepSeek-V4.1-Flash-1017. It reads images. The old names
deepseek-v4-flashanddeepseek-v4-flash-vision-expstill work: those models were retired, and DeepSeek serves and bills those requests as Flash. - deepseek-v4-pro runs DeepSeek-V4-Pro-0813. It doesn't take images.
Both have a 1M-token context, up to 384K tokens of output, and thinking and non-thinking modes (thinking is the default). Both support JSON output, tool calls, the Responses API and the Anthropic API format.
Peak and off-peak hours
Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, not counting Chinese public holidays. That is 35 hours a week; the other 133 hours, including all weekends, are billed at half price. In other time zones (in November, without daylight saving time) the peak hours are:
| Time zone | Peak hours |
|---|---|
| UTC / London | 01:00-04:00 and 06:00-10:00, Mon-Fri |
| Beijing (UTC+8) | 09:00-12:00 and 14:00-18:00, Mon-Fri |
| Central Europe (UTC+1) | 02:00-05:00 and 07:00-11:00, Mon-Fri |
| India (UTC+5:30) | 06:30-09:30 and 11:30-15:30, Mon-Fri |
| US Eastern (UTC-5) | 20:00-23:00 Sun-Thu, and 01:00-05:00 Mon-Fri |
| US Pacific (UTC-8) | 17:00-20:00 Sun-Thu, and 22:00-02:00 overnight Sun-Thu |
In the Americas the expensive hours fall in the evening and at night, so daytime use there is almost all off-peak. If you run batch jobs, schedule them outside these windows.
Cost calculator
The calculator uses peak-hour prices, the most you would pay. Off-peak costs are half.
| Model | Input / 1M | Output / 1M | Cached / 1M | Per request | Per month |
|---|---|---|---|---|---|
DeepSeek V4.1 Flashdeepseek-flash$0.30 in / $1.20 out per 1M | $0.30 | $1.20 | $0.006 | $0.0012 | $1.80 |
DeepSeek V4 Prodeepseek-v4-pro$1.32 in / $3.96 out per 1M | $1.32 | $3.96 | $0.044 | $0.0046 | $6.93 |
USD list prices per 1M tokens, standard tier, checked on on each provider's pricing page. A month is 30 days. Taxes, batch discounts, tool fees (web search) and cache-write charges are left out.
DeepSeek V4.1 Flash, DeepSeek V4 Pro: Peak-hour price. Off-peak (most of the week) is half.
What a request costs
- A chat message of about 2,000 tokens in and 500 out costs $0.0012 on deepseek-flash at peak and $0.0006 off-peak. At off-peak rates, $2 covers more than 3,000 messages. The same message on deepseek-v4-pro costs about $0.0046 at peak.
- A coding agent step with 40,000 tokens of context (80% of it from the cache) and 1,500 tokens of output costs about $0.0044 on deepseek-flash and $0.018 on deepseek-v4-pro at peak. The same step costs about $0.034 on Claude Sonnet 5.5 (see Claude API pricing).
Cache hits keep agent costs low. On deepseek-flash a cached input token costs a fiftieth of an uncached one.
Other DeepSeek APIs
| API | Price |
|---|---|
| Search API (text) | $0.75 per 1,000 successful requests |
| Fetch API | $0.75 per 1,000 successful requests |
| File parsing and OCR | $0.225 per 1,000 pages |
| Text to speech | $0.45 per 10,000 characters (limited-time price $0.225) |
| Speech to text | $0.116 per hour of audio (limited-time price $0.058) |
Since 12:00 Beijing time on 2026-11-18, search and web fetch that a model triggers on its own during a request (through DeepSeek's web tools, the web_search tool in the Anthropic format or in the Responses API) are no longer billed separately. You pay only for the tokens.
How billing works
- You top up a balance on the DeepSeek platform and each request is deducted from it. A granted (promotional) balance is used first.
- The pricing page lists no monthly fee and no free tier.
- DeepSeek lists concurrency limits of 2,500 requests for deepseek-flash and 500 for deepseek-v4-pro, well above what one person's apps need.
- DeepSeek says it may change prices and recommends topping up based on your actual usage.
To create a key and connect it to an app, see how to get a DeepSeek API key. To compare DeepSeek with Claude, GPT, Gemini and Grok in one table, see the LLM API price comparison.
Apps that work with your DeepSeek API key
Questions
How much does the DeepSeek API cost?
On 2026-11-18, deepseek-flash cost $0.30 per million input tokens and $1.20 per million output tokens at peak hours, and half that off-peak ($0.15 / $0.60). deepseek-v4-pro cost $1.32 / $3.96 at peak and $0.66 / $1.98 off-peak.
When are DeepSeek's off-peak hours?
Everything outside 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday. Weekends and Chinese public holidays are off-peak all day.
Is the DeepSeek API free?
No. DeepSeek bills a prepaid balance. If you have a granted (promotional) balance, it is used before the money you topped up.
What is a cache hit?
When the start of your prompt matches a recent request, DeepSeek reads it from cache and bills it at the cache-hit price: $0.006 per million tokens for deepseek-flash at peak instead of $0.30.
Do I pay extra for web search?
The standalone Search and Fetch APIs cost $0.75 per 1,000 successful requests. Since 12:00 Beijing time on 2026-11-18, search and web fetch that the model triggers during a chat request are no longer billed; you pay only for tokens.
Sources
Checked against the providers' own documentation on 2026-11-18. Prices and limits change; the links below have the current figures.