You Can Use the Uncensored Version of DeepSeek Through Perplexity
If you’ve been thinking about trying DeepSeek, a popular new AI app from China, but have been put off by concerns about privacy and censorship , you’ll be interested to know that a modified version of the DeepSeek R1 model is now available inside Perplexity.
Perplexity has long given users a choice of first-party and third-party AI to answer queries, and DeepSeek R1 is the latest to join the list. Perplexity co-founder Aravind Srinivas promised that censorship fences have been lifted (so yes, you can ask about Tiananmen Square) and that all related data (like your tips and answers to them) is stored in the US .
All this is possible because DeepSeek models are open source: anyone can use and adapt them for free. If you have the technical know-how, you can modify these AI models yourself, although most of us will want experts (like the Perplexity team) to take care of this. If you’re interested, there’s more about it here .
Most notably, Perplexity adds the DeepSeek R1 reasoning model rather than the more generic V3 model, and both Pro subscribers and those who use Perplexity for free can access these models: you get five queries per day for free or 500 per day if you pay $20 per month.
If you want to see what the fuss is all about when it comes to DeepSeek speed and performance without worrying about your data being passed back to the Chinese government or that you’ll face China-centric censorship, this is one of the best ways to do it . Your data is still stored by Perplexity, but in the United States and in accordance with the company’s privacy policy .
It’s also worth noting that other AI providers are following Perplexity’s lead: Microsoft is bringing DeepSeek R1 to Windows as part of Copilot, and it is now also featured on You.com’s AI platform , available alongside other models such as Anthropic, Meta. , Grok and OpenAI.
Using DeepSeek inside Perplexity
You can use Perplexity on the web , Android and iOS . On the web, your prompts can be entered into the main text box on the Home tab or by clicking New Topic on the left. To get an answer from DeepSeek R1, turn on the Pro toggle, then select “Reasoning with R1” from the drop-down menu.
Everything works similarly with mobile applications. Tap the Ask Something box on the main Perplexity tab (the magnifying glass icon in the bottom left corner), then tap the Pro shortcut and select Reasoning with R1 . You can then enter and submit your request to the AI.
Because DeepSeek R1 is a reasoning model, you’ll see what it thinks about your clues and explain how it thinks as it goes along. Answers are accompanied by links to websites where appropriate, and at the end of the answer you receive icons for copying, sharing, and re-creating the answer Perplexity and DeepSeek gave you.
I tested the Perplexity version of DeepSeek by asking a few questions that might have different answers if asked in China compared to the rest of the world (questions about Taiwan, for example), and there was a free and complete flow of information. in return. When calling up the same prompts on DeepSeek itself, the following message appeared: “Sorry, this is beyond my current capabilities. Let’s talk about something else.”
As for DeepSeek R1’s reasoning, based on the time I spent with it, it seems pretty convincing – it definitely makes the AI answers more trustworthy if you can see how they were arrived at and the web sources consulted at the same time. way. You can view all of this after the answers have been submitted, as well as while they are being generated.
We’ve previously written about how well Perplexity stacks up against competitors like Gemini and ChatGPT by combining both generative AI and web search, while also giving you a choice of different AI models. The company also just launched Perplexity Assistant on Android, so you can use an AI bot instead of Gemini or Bixby.