DeepSeek API pricing guide & calculator
DeepSeek API prices range from $0.30 to $1.32 per million input tokens and $1.20 to $3.96 per million output tokens, depending on the model. Below: every current model’s price, what common jobs cost per month, and a calculator for your own usage.
Current DeepSeek API prices
| Model | Input | Cached input | Output | Batch in / out |
|---|---|---|---|---|
| DeepSeek V4 Proflagship · 1M tokens | $1.32 | $0.044 | $3.96 | — |
| DeepSeek Flash (V4.1)fast · 1M tokens | $0.30 | $0.006 | $1.20 | — |
Per million tokens, standard tier, in US dollars. Checked October 10, 2026 against DeepSeek’s official pricing. “Cached input” is the price for text you reuse across requests, like long instructions.
- DeepSeek V4 Pro: Peak-hour price. Off-peak (most hours) is half: $0.66 / $1.98.
- DeepSeek Flash (V4.1): Peak-hour price. Off-peak is half: $0.15 / $0.60.
Estimate your monthly DeepSeek API cost
Pick a job and how many you’ll run a month. DeepSeek’s models are compared with OpenAI and Gemini and Claude; tap a provider to add or remove it. How the estimates work.
| Model | Per ticket | Per month | vs. cheapest |
|---|---|---|---|
| Claude Haiku 5.5Anthropic · fast | $0.00042 | $0.85 | 1× |
| GPT-6 LunaOpenAI · fast | $0.00042 | $0.85 | 1× |
| DeepSeek Flash (V4.1)DeepSeek · fast | $0.0012 | $2.34 | 2.8× |
| Gemini 3.5 Flash-LiteGoogle · fast | $0.0016 | $3.25 | 3.8× |
| Gemini 3.8 FlashGoogle · balanced | $0.0032 | $6.38 | 7.5× |
| DeepSeek V4 ProDeepSeek · flagship | $0.0047 | $9.37 | 11× |
| Claude Sonnet 5.5Anthropic · balanced | $0.0085 | $17.00 | 20× |
| GPT-6.1 SolOpenAI · balanced | $0.0085 | $17.00 | 20× |
| Gemini 3.1 Pro (Preview)Google · flagship | $0.0092 | $18.40 | 22× |
| Claude Opus 5.5Anthropic · flagship | $0.0170 | $34.00 | 40× |
| Claude Fable 5.1Anthropic · flagship | $0.0425 | $85.00 | 100× |
| GPT-6 AstraOpenAI · flagship | $0.0425 | $85.00 | 100× |
Published list prices checked October 10, 2026, before tax. Token counts are estimates and vary by provider, since each counts text a little differently. Subscriptions like Claude Pro, ChatGPT Plus, or Google AI Pro (about $20 a month) are for using the chat apps yourself and can’t be called from your own software. Calculated in your browser; nothing is saved.
Which DeepSeek model should you use?
DeepSeek keeps things simple, with two current models. DeepSeek V4 Pro is the capable one, at $1.32 per million input tokens and $3.96 per million output tokens during peak hours. DeepSeek Flash (V4.1) is the fast, low-cost model at $0.30 / $1.20. Both have a 1 million token context window and allow very long outputs.
The headline feature is time-of-day pricing. DeepSeek’s peak hours are 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays. At every other time, prices are half. For a business in the Americas, most of the working day falls outside those windows, so many requests are billed at the lower, off-peak rate. The calculator uses peak prices by default; switch on “DeepSeek off-peak” to see the difference.
Worked example: scheduling bulk work off-peak
Suppose you summarize 300 long documents a month and answer 2,000 support tickets. At peak prices on V4 Pro that’s about $12.98 a month. Run the summaries as an overnight job outside the peak windows and they cost half, and if your support traffic mostly arrives during the Americas’ working day, most of it is billed off-peak too, bringing the total close to $6.49. For comparison, the same work costs about $23.3 on Claude Sonnet 5.5 or GPT-6.1 Sol at standard prices. The catch is that you have to know when your requests run; if you can’t control timing, budget at the peak price.
Its cache pricing is also unusually low: a cache hit costs $0.044 per million tokens on V4 Pro and $0.006 on Flash at peak, so apps that resend the same instructions or documents get very cheap repeated input.
What common jobs cost on DeepSeek
Monthly cost at standard prices, with no caching or batch discounts, for typical volumes. The token sizes behind each job are explained in how we calculate.
| Job (per month) | V4 Pro | Flash (V4.1) |
|---|---|---|
| 1,000 × draft an email | $2.64 | $0.72 |
| 2,000 × answer a support ticket | $9.37 | $2.34 |
| 300 × summarize a 10-page document | $3.6 | $0.882 |
| 100 × write a 1,500-word article | $0.95 | $0.276 |
| 5,000 × write a product description | $8.18 | $2.28 |
| 3,000 × chatbot conversation (10 turns) | $71.28 | $18 |
| 200 × ai coding or research agent task | $47.52 | $11.4 |
At peak prices, answering 2,000 support tickets a month costs about $9.37 on V4 Pro and $2.34 on Flash; off-peak, half that. V4 Pro is priced well below the top models from OpenAI and Anthropic, and even below their middle-tier models.
How DeepSeek compares on price
DeepSeek doesn’t have a model in the “balanced” slot the big three use, so the table compares the big three’s everyday models. DeepSeek V4 Pro at peak ($9.37 for 2,000 tickets) costs less than Claude Sonnet 5.5 or GPT-6.1 Sol but more than Gemini 3.8 Flash; off-peak, it costs less than Gemini Flash too. Use the calculator above to compare DeepSeek Flash with the budget models.
| Balanced model | Input / output per 1M | 2,000 tickets |
|---|---|---|
| Gemini 3.8 FlashGoogle | $0.75 / $3.75 | $6.38 |
| Claude Sonnet 5.5Anthropic | $2 / $10 | $17 |
| GPT-6.1 SolOpenAI | $2 / $10 | $17 |
Discounts: batch, caching, and free use
Off-peak: 50% off. DeepSeek’s discount isn’t a separate batch product. Every request made outside the peak windows is simply billed at half price, automatically. If you can schedule bulk work, such as overnight in Asia or during the Americas’ working day, you get the discount without changing anything else.
Cache hits: around 97–98% off. When input repeats across requests, DeepSeek bills the repeated part at its cache-hit price, a small fraction of the normal input price.
There’s no batch tier listed, and no priority tier.
Free tier: No free API tier; the chat app is free.
What surprises people on the bill
Your cost depends on when requests run
Because the same request costs twice as much during peak hours, monthly bills can vary with when your users are active. If most of your traffic falls in the peak windows (early morning UTC on weekdays), budget at peak prices.
Data location and compliance
DeepSeek is a Chinese company, and its first-party API is operated from China. Some businesses, especially in regulated industries or government work, have policies about where data may be processed. Check yours before sending customer or confidential data. Some cloud providers host DeepSeek’s open models in other regions at their own prices.
Fewer extras than the big three
DeepSeek offers a lean API. If you need built-in web search, file handling, or enterprise features like regional processing and detailed admin controls, compare what each provider includes, not just the token price.
Long outputs are allowed, and billed
Both models allow up to 384,000 output tokens. That’s useful for long documents, but a runaway response is expensive, so set a sensible maximum output length in your app.
DeepSeek API vs. DeepSeek chat app
DeepSeek’s chat app is free, so there’s no paid consumer subscription to compare against. The API is for building DeepSeek into your own software, billed per token.
For a sense of scale, $20 of API usage buys about 7,575 email drafts on V4 Pro at peak prices, or about 27,777 on Flash, using our token estimates, and roughly double that off-peak.
| DeepSeek subscription | Price |
|---|---|
| DeepSeek chat app | Free (free) |
Who the DeepSeek API is a good fit for
DeepSeek is a strong option for cost-sensitive, high-volume work where you can control timing, such as bulk summarization, data processing, and internal tools, and where your data policies allow it. Its very cheap cache hits suit apps that reuse the same long instructions.
For customer-facing products, regulated data, or teams that need enterprise controls, the bigger providers may be worth the extra cost. Test quality on your own tasks: a cheaper model that needs more retries or more human review can end up costing more overall.
Frequently asked questions
How much does the DeepSeek API cost?
At peak hours, DeepSeek V4 Pro costs $1.32 per million input tokens and $3.96 per million output tokens, and DeepSeek Flash costs $0.30 / $1.20. Outside peak hours, both are half price. Cache hits cost a small fraction of normal input.
When are DeepSeek’s off-peak hours?
DeepSeek’s peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other times, including weekends, are off-peak, at 50% of peak prices.
Is the DeepSeek API cheaper than OpenAI?
At list price, yes, compared with OpenAI’s top and middle models: DeepSeek V4 Pro costs less than GPT-6.1 Sol and much less than GPT-6 Astra, especially off-peak. OpenAI’s GPT-6 Luna is cheaper than DeepSeek Flash for short prompts at peak hours.
Is the DeepSeek API free?
No. The API is paid per token, although the DeepSeek chat app is free to use.
Is it safe to use DeepSeek for business data?
That depends on your data and your policies. DeepSeek’s own API is operated from China. Review its privacy policy and your industry’s rules before sending customer or confidential information, and consider a provider or hosting option that meets your requirements.