Chat & AssistantsFreemium
DeepSeek Pricing: What the API Costs and What the Chat App Doesn't
DeepSeek publishes no subscription plans. The chat app has no listed price, and the API is pay-as-you-go: deepseek-flash costs $0.15 per million input tokens and $0.60 per million output tokens off-peak, with rates doubling in peak hours. Data is stored in China.
Pricing checked · Page updated
DeepSeek pricing
From DeepSeek's own pricing page, read on November 20, 2026. Prices change; confirm on the tool's site before you buy.
| Plan | Price | What you get |
|---|---|---|
| Chat app | No listed price | Web and mobile chat with an account. DeepSeek publishes no subscription tiers and no usage limits for it. |
| API: deepseek-flash (off-peak) | $0.15 per 1M input tokens | $0.003 per 1M cached input tokens and $0.60 per 1M output tokens. 1M context, 384K maximum output. |
| API: deepseek-flash (peak) | $0.30 per 1M input tokens | $0.006 per 1M cached input tokens and $1.20 per 1M output tokens during peak hours. |
| API: deepseek-v4-pro (off-peak) | $0.66 per 1M input tokens | $0.022 per 1M cached input tokens and $1.98 per 1M output tokens. 1M context, 384K maximum output. |
| API: deepseek-v4-pro (peak) | $1.32 per 1M input tokens | $0.044 per 1M cached input tokens and $3.96 per 1M output tokens during peak hours. |
Key features
- Chat app on web and mobile with file uploads and shareable chats
- Two API models, deepseek-flash and deepseek-v4-pro, each with a 1M token context
- Up to 384K tokens of output per request
- Cache-hit pricing for repeated input
- Off-peak rates at half the peak price
- File parsing and OCR, text to speech and speech to text through the API
Two different products share the DeepSeek name, and only one of them has a price list. Searches for DeepSeek pricing usually come from people who expect a Plus or Pro plan like the ones ChatGPT and Claude sell. There isn't one. The chat app has no published price at all, and the developer API is billed by the token with no monthly fee.
Is DeepSeek free?
The chat app, on the web at chat.deepseek.com and on mobile, asks you to create an account and nothing more. DeepSeek's site lists no subscription, no upgrade and no paid tier for it. In practice that makes it a no-cost product, but DeepSeek does not publish any usage limits for the app, so there is no written promise about how much you can use it.
The API is separate and is not free. You top up a balance and each request draws it down. DeepSeek's pricing page mentions that a granted balance, when an account has one, is used before the topped-up balance, but it does not promise free credits to new accounts.
DeepSeek pricing for the API
DeepSeek sells two models through the API, and each has a peak and an off-peak rate. All prices are per one million tokens.
deepseek-flash (currently DeepSeek-V4.1-Flash) is the cheap, fast model:
- Input, cache hit: $0.003 off-peak, $0.006 peak
- Input, cache miss: $0.15 off-peak, $0.30 peak
- Output: $0.60 off-peak, $1.20 peak
deepseek-v4-pro is the larger model:
- Input, cache hit: $0.022 off-peak, $0.044 peak
- Input, cache miss: $0.66 off-peak, $1.32 peak
- Output: $1.98 off-peak, $3.96 peak
Both models have a 1M token context length and a maximum output of 384K tokens.
Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, excluding Chinese public holidays. Every other hour, including all weekend, is off-peak and half the price. For a US team that is good news: the peak windows fall in the evening and overnight on the East Coast, so a normal American working day is billed at the lower rate.
A few extras are priced separately. File parsing and OCR cost $0.225 per 1,000 pages. Text to speech is listed at $0.45 per 10,000 characters and speech to text at $0.116 per hour of audio, both with a limited-time 50% discount at the time of checking. Since November 18, 2026, search and fetch tool calls triggered during a model call are no longer billed per request; their results count as input tokens instead.
How billing works
There are no tiers. DeepSeek's help center says there is "a unified pricing standard and no tiered plans", and that higher rate limits are granted on request at no additional cost. You can top up by PayPal, bank card, Alipay or WeChat Pay.
Two details are friendlier than usual. Top-up balances do not expire, and unused balance can be refunded from the billing page. Money you load and do not spend is not lost.
The cache-hit price matters more than it looks. If your application sends the same long system prompt or the same document with every request, the repeated part is billed at $0.003 per million tokens on deepseek-flash instead of $0.15. For chat-style products with long fixed instructions, most of the input bill disappears.
Privacy and where your data goes
This is the part to read before signing up. DeepSeek's privacy policy states that it collects, processes and stores personal data in the People's Republic of China. It also says user data may be used to train and improve its models. You have the right to opt out of training, but the route given is an email to DeepSeek's privacy address, not a switch in settings.
The terms of use are governed by the laws of mainland China, and say the service is primarily intended for adults; under-18s need a guardian's consent.
None of this makes the product unusable. It does mean you should not paste client data, health information or unreleased code into the chat app unless your organization has approved it, and that regulated businesses will probably rule it out.
Who it suits, and who should look elsewhere
DeepSeek fits developers whose main constraint is cost. Batch jobs, classification, summarizing large volumes of text and prototypes all benefit from sub-dollar output pricing, and the off-peak schedule rewards work that can run at any hour.
It also fits individuals who want a capable assistant without paying a subscription and who are comfortable with the data terms.
It does not fit teams that need data kept in the US or EU, published uptime or usage commitments for the chat app, or the extra tools that the subscription assistants bundle, such as image generation and coding agents. If the chat app slows down or limits you, there is no paid plan to move to.
Alternatives
ChatGPT has unlimited free text chat and a $20 Plus plan with far more built-in tools. Claude's free plan includes web search and Artifacts, with Pro at $20 a month. Gemini starts at $4.99 a month and bundles Google storage. Venice AI is the one to look at if privacy is the reason you are shopping around.
Pros and cons
What it does well
- No subscription: the API is pay-as-you-go and top-up balances do not expire
- Unused API balance can be refunded
- Cached input costs a fraction of fresh input, from $0.003 per million tokens
- One price list for everyone, with rate-limit increases granted at no extra cost
Where it falls short
- Personal data is collected, processed and stored in the People's Republic of China
- Chats can be used to train DeepSeek's models unless you opt out by email
- Peak-hour API calls cost twice the off-peak rate
- No published usage limits for the chat app, and no paid tier to buy your way past them
DeepSeek alternatives

ChatGPT
The default AI assistant: unlimited free text chat, Plus at $20/month, and Pro tiers up to $500.
Freemiumfrom $20/month
Claude
Anthropic's assistant: a capped free plan, Pro at $20/month, and Max and Ultra tiers for heavy use.
Freemiumfrom $20/month
Gemini
Google's assistant, sold as a bundle: AI plans from $4.99/month that also include Google storage.
Freemiumfrom $4.99/month
Venice AI
Private, uncensored AI chat and image generation with a small free tier and an $18 Pro plan.
Freemiumfrom $18/month
DeepSeek FAQ
Is DeepSeek free?
The chat app has no listed price: you create an account and use it, and DeepSeek sells no subscription for it. It does not publish usage limits for the app. The API is not free and is billed per token from a prepaid balance.
How much does the DeepSeek API cost?
DeepSeek pricing is per million tokens. deepseek-flash costs $0.15 input and $0.60 output off-peak; deepseek-v4-pro costs $0.66 input and $1.98 output off-peak. Peak-hour rates are double, and cached input is far cheaper, from $0.003.
Is DeepSeek safe to use?
DeepSeek's privacy policy says personal data is stored in the People's Republic of China and may be used to train its models, with an opt-out available by email. Avoid entering confidential or regulated information unless your organization allows it.
Does DeepSeek API credit expire?
No. DeepSeek's help center states that top-up balances do not expire, and unused balance can be refunded through the billing page.
What is the best DeepSeek alternative?
ChatGPT and Claude are the main subscription assistants, both with free plans. Gemini is the cheapest paid option at $4.99/month, and Venice AI is aimed at people who put privacy first.