| What the sources support |
|---|
| What not to assume |
|---|
| Web chat | Someone opening DeepSeek in a browser | DeepSeek’s homepage links to “Start Now” with “Free access to DeepSeek.” | Free access does not prove unlimited usage or permanent terms. |
| Mobile app | Someone using DeepSeek on a phone | The Google Play listing describes DeepSeek’s official AI assistant as free to use. | Heavy or work-related users should still check current in-app limits. |
| API | Developers connecting DeepSeek to software or workflows | DeepSeek’s homepage separates “Access API” from the chat route, and the API pricing source lists token-based rates. | Do not treat the API as automatically free just because the chat app is free. |
DeepSeek’s homepage points users to a chat experience with the line “Free access to DeepSeek.” Its Google Play listing also describes DeepSeek as the official AI assistant that can be used for free.
That supports a simple conclusion for everyday use: if you want to chat with DeepSeek directly in a browser or mobile app, you can start with the free access described in those listings.
What the supplied sources do not establish is that every feature, model, usage level, or policy is unlimited. For casual use, that may not matter. For work, high-volume research, or anything mission-critical, check the live app or account terms before building a routine around it.
The API is the product you use when you want software — not just a person in a chat window — to call DeepSeek. That might mean connecting it to an app, internal tool, customer support bot, website feature, or automated workflow.
DeepSeek’s homepage makes the split visible: “Start Now” leads to the chat experience, while “Access API” is a separate entry point for building with the models. The API pricing source also says the app/web experience may be presented separately from API billing.
For API use, the listed public rates are per 1 million tokens: $0.028 for input cache hits, $0.28 for input cache misses, and $0.42 for output tokens.
In plain English, a token is a unit of text processed by the model. Your API cost depends on the amount of input text, the amount of output text, and whether input is charged at the cache-hit or cache-miss rate.
For API planning, the practical unit is not a chat message. A short question with a short answer uses far fewer tokens than a long document analysis or an automated workflow that sends large prompts repeatedly.
Because the pricing source bills by tokens used, the budget question is: how many input and output tokens will your integration process?
Some third-party pages describe API free tiers, but their numbers do not line up. Price Per Token says new accounts get 5 million free tokens, while MyDeepSeekAPI says DeepSeek V3 and R1 include 500 million tokens per month.
Those are very different claims. Before you count on free API usage in a production plan, verify the current allowance in the place you sign up or in your API dashboard. A safer approach is to treat any free API allowance as trial credit and budget ongoing use against the published token rates.
DeepSeek is reasonably described as free to access on web and app, based on the homepage and Google Play listing. But that does not make the API free by default. For integrations, the pricing source lists rates per 1 million tokens for input cache hits, input cache misses, and output tokens.
Rule of thumb: chat with DeepSeek directly and you can start free; build with the DeepSeek API and you need to budget by token.