DeepSeek Pricing (2026): Free App and API Costs Explained
- Made by
- Hangzhou DeepSeek Artificial Intelligence Co., Ltd.
- App price
- Free, with no paid tier
- Platforms
- Web, iOS, Android, API
- API from
- $0.15 in / $0.60 out per 1M tokens (Flash, off-peak)
- Off-peak discount
- 50%, since 16 August 2026
- Context window
- 1M tokens (API)
- Data stored
- People's Republic of China
- Open weights
- Yes, MIT licence (V4)
The DeepSeek app is completely free, and there is no paid plan to upgrade to. You get the chatbot on the web, iOS and Android at no cost, and the App Store listing shows no in-app purchases.
Developers pay per token for the API, and it is one of the cheapest good APIs available. The fast model, deepseek-flash (DeepSeek-V4.1-Flash), costs $0.30 per million input tokens and $1.20 per million output tokens at peak times. The stronger deepseek-v4-pro costs $1.32 and $3.96. Since 16 August 2026, every rate is half price off-peak, which covers all weekends and most of each weekday. Repeated prompt text that hits DeepSeek's cache costs 30 to 50 times less than new input.
The real price is your data. DeepSeek's privacy policy says it stores personal data in China and may use your inputs to train its models unless you opt out. Keep private, client and company information out of it, or run DeepSeek's open-weight models somewhere else.
Prices checked 25 September 2026 on DeepSeek's official pricing page. Prices are in US dollars unless stated and can change at any time.
DeepSeek plans and prices
DeepSeek app
- Chat on chat.deepseek.com, iOS and Android
- Instant Mode for quick answers
- Expert Mode for step-by-step reasoning (runs V4-Pro since 13 August 2026)
- Web search
- No image generation and few extras
- Data stored in China; chats may train DeepSeek's models unless you opt out
API: deepseek-flash (V4.1-Flash)
- Cache-hit input: $0.006 per 1M tokens peak, $0.003 off-peak
- 1M-token context window and up to 384K output tokens
- Image input (vision), tool calls and JSON output
- Thinking and non-thinking modes
- OpenAI and Anthropic API formats
- Concurrency limit of 2,500 requests
API: deepseek-v4-pro (V4-Pro)
- Cache-hit input: $0.044 per 1M tokens peak, $0.022 off-peak
- 1M-token context window and up to 384K output tokens
- Tool calls, JSON output and adjustable reasoning effort
- No image input
- Concurrency limit of 500 requests
Open weights (self-hosted)
- DeepSeek V4 model weights on Hugging Face
- Commercial use allowed under the MIT licence
- V4-Pro has 1.6 trillion parameters, so running it needs a large GPU cluster
Is there a free plan?
Yes. The whole DeepSeek app is free. There is no subscription, and the US App Store listing shows no in-app purchases. You get Instant Mode for quick replies and Expert Mode, which has run DeepSeek's strongest model, V4-Pro, since 13 August 2026. It is strong at maths, code and step-by-step reasoning.
What you do not get: image generation, advanced voice, agents or the polish of paid rivals. App Store reviewers give it 3.9 out of 5 from about 12,000 ratings.
The catch is privacy. DeepSeek's privacy policy (updated 10 February 2026) says it collects, processes and stores personal data in the People's Republic of China, that your inputs may be used to train its models, and that you can opt out by emailing privacy@deepseek.com. The API is a separate, paid product with no free tier listed on the pricing page.
What you will actually pay
Worked examples using the published prices above.
| Scenario | Cost | How we got there |
|---|---|---|
| Personal use of the DeepSeek app | $0 | The app has no paid tier. |
| Small app on Flash: 1,000 requests a day, each 2,000 tokens in and 500 out | $18/month off-peak; $36/month at peak | Per day: 2M input × $0.15 + 0.5M output × $0.60 = $0.60 off-peak, or 2M × $0.30 + 0.5M × $1.20 = $1.20 at peak. Times 30 days. |
| Coding agent on V4-Pro: 50M input tokens a month (80% cache hits) and 5M output | $34.76/month at peak; $17.38 off-peak | Peak: 40M × $0.044 + 10M × $1.32 + 5M × $3.96 = $1.76 + $13.20 + $19.80. With no cache hits the same month would cost 50M × $1.32 + 5M × $3.96 = $85.80. |
DeepSeek pricing compared with rivals
| Tool | Paid plans from | Free plan | Note |
|---|---|---|---|
| DeepSeek | Free, with no paid tier | See above | The tool on this page |
| Qwen Chat | Free | Yes: the chat app has no paid consumer tier | Alibaba's free chatbot with image generation; API sold separately. |
| Kimi | $19/month (Moderato); $15/month billed yearly | Yes: basic chat | Paid tiers up to $199 a month add agent credits; API billed separately. |
| ChatGPT | $8/month (Go, US) | Yes, with image creation and voice | More tools on the free plan than DeepSeek; Plus is $20. |
| Gemini | $4.99/month (Google AI Plus, US) | Yes: Flash model, Deep Research, image generation | API: Gemini 3.1 Flash-Lite costs $0.25 in / $1.50 out per 1M tokens. |
| Mistral Vibe (formerly Le Chat) | $14.99/month (Pro) | Yes, with limited messages | API: Mistral Large costs $0.50 in / $1.50 out per 1M tokens. EU-based. |
| Grok | $10/month (SuperGrok Lite) | Yes, with tight limits | API: Grok 4.3 costs $1.25 in / $2.50 out; Grok 4.7 $2 / $6. |
The app is worth using for free, for the right tasks. Expert Mode gives you DeepSeek's strongest model at no cost, which makes it a good homework, maths and coding helper. It is not worth the privacy trade for anything personal, confidential or work-related.
The API is some of the best value in AI. Off-peak, Flash costs $0.15 in and $0.60 out per million tokens, cheaper than Gemini 3.1 Flash-Lite ($0.25 and $1.50) and far cheaper than Grok 4.7 ($2 and $6). V4-Pro at $1.32 and $3.96 at peak is a low price for a model of its strength.
- Pick Flash for chatbots, summaries and agents that run all day.
- Pick V4-Pro for harder coding and reasoning, and schedule big batch jobs off-peak.
- Pick the open weights on another host if you like the models but cannot send data to China.
See our DeepSeek V4 page for benchmarks and DeepSeek vs ChatGPT for a head-to-head.
DeepSeek API pricing in detail
The API is billed per token (a token is roughly a word; DeepSeek estimates 1 English character at about 0.3 tokens). You top up a prepaid balance and DeepSeek deducts fees from it, using any granted balance first.
| Price per 1M tokens | 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 |
From DeepSeek's pricing page on 25 September 2026.
When is off-peak? 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 is off-peak, including whole weekends and Chinese public holidays. DeepSeek introduced this split on 16 August 2026.
Worked example: 1,000 requests, each with 2,000 new input tokens and 500 output tokens, is 2 million input and 0.5 million output tokens. On Flash that costs (2 × $0.30) + (0.5 × $1.20) = $1.20 at peak and $0.60 off-peak. On V4-Pro it costs $4.62 at peak and $2.31 off-peak.
Privacy and data location
DeepSeek is free in money, not in data. Its privacy policy, updated 10 February 2026, says:
- Who holds your data: Hangzhou DeepSeek Artificial Intelligence Co., Ltd., registered in China.
- Where: personal data is collected, processed and stored in the People's Republic of China.
- Training: your inputs may be used to train its models. You can opt out by emailing privacy@deepseek.com.
- Authorities: DeepSeek may share data with law enforcement and public authorities where needed to comply with law.
The iOS app's privacy label lists contact details, user content, coarse location, search history and device identifiers as data linked to you.
Regulators have acted on this. Italy's data protection authority blocked DeepSeek on 30 January 2025, and in June 2025 Berlin's data protection commissioner asked Apple and Google to remove the app in Germany, saying its transfer of German users' data to China was unlawful. Many government bodies also ban it on official devices.
The workaround: DeepSeek's V4 models are open weights under the MIT licence. Running them yourself, or through a hosting provider in your own region, gives you the same models without sending prompts to DeepSeek.
- Schedule batch jobs for weekends or outside 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays. Off-peak rates are exactly half.
- Keep the start of your prompts identical between calls (system prompt, instructions, shared documents) so they hit the cache. Cached input on Flash costs $0.006 per million tokens instead of $0.30.
- Use Flash by default and send only the hard problems to V4-Pro. Flash costs less than a quarter of V4-Pro per token.
- Thinking mode is on by default. Turn it off for simple tasks such as classification or short rewrites to get faster, shorter replies.
- In the free app, never paste passwords, client files, medical details or ID numbers. If you want the models for private work, use the open weights on a host in your own region.
Jargon explained
- Token
- The small chunk of text an AI model reads and writes, often a word or part of one. API prices are quoted per million tokens.
- Cache hit
- When the start of your prompt matches text DeepSeek processed recently, so that part is billed at a much lower rate.
- Off-peak pricing
- DeepSeek's half-price API rate, which applies at weekends and outside two blocks of weekday hours.
- Open weights
- Model files anyone can download and run. DeepSeek publishes its V4 models this way under the MIT licence.
- Data jurisdiction
- The country whose laws apply to your stored data, including rules on when authorities can access it.
Frequently asked questions
Is DeepSeek free?
Yes. The DeepSeek app on the web, iOS and Android is free and has no paid tier. Only the developer API costs money.
How much does the DeepSeek API cost?
At peak times, deepseek-flash costs $0.30 per million input tokens and $1.20 per million output tokens, and deepseek-v4-pro costs $1.32 and $3.96. Off-peak rates are half. Cached input is much cheaper: $0.006 on Flash and $0.044 on V4-Pro at peak. Prices checked 25 September 2026.
When are DeepSeek's off-peak hours?
All hours except 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays. Weekends and Chinese public holidays are off-peak all day.
Is there a DeepSeek Pro or paid subscription?
No. Unlike ChatGPT or Gemini, DeepSeek sells no consumer subscription. "V4-Pro" is a model name, and you can use it free in the app's Expert Mode or pay for it through the API.
Where does DeepSeek store my data?
In China. DeepSeek's privacy policy says it collects, processes and stores personal data in the People's Republic of China and may use inputs for training unless you opt out.
What is a cache hit in DeepSeek pricing?
When the start of your prompt matches text DeepSeek has recently processed, that part is billed at the cheaper cache-hit rate. On Flash that is $0.006 per million tokens at peak instead of $0.30.
Sources
Every fact on this page comes from public information. Vendor figures are labelled as vendor claims.
- Models and pricing (DeepSeek)
- DeepSeek-V4-Pro GA release note (DeepSeek)
- DeepSeek-V4 preview release note (DeepSeek)
- Thinking mode (DeepSeek)
- Token and token usage (DeepSeek)
- DeepSeek privacy policy (DeepSeek)
- DeepSeek - AI Assistant on the App Store (Apple)
- DeepSeek-V4-Pro model card (Hugging Face)
- DeepSeek AI blocked by Italian authorities as other member states open probes (Euronews)
- Germany tells Apple, Google to remove DeepSeek from the country's app stores (TechCrunch)
- Gemini API pricing (Google)
- Mistral pricing (Mistral AI)
- Grok models and pricing (SpaceXAI Docs)
- Kimi membership pricing (Moonshot AI)
- Qwen Studio (Alibaba)
- ChatGPT pricing (OpenAI)
- Google AI plans (Google)
- Claude pricing (Anthropic)
- Introducing Meta One (Meta)