deepseek ai

DeepSeek AI: The Free Chatbot That Does It All

Have you heard about DeepSeek? It’s this crazy AI tool that’s kinda like ChatGPT but with something extra. And the best part? It’s completely free right now! So, DeepSeek is an AI chatbot that can do all the usual stuff—answer questions, write emails, help with coding, generate content, solve homework, and even simplify complex topics like you’re five years old. Let me give you all the details.

It’s like having a personal AI assistant that’s super smart, privacy-focused, and totally free (at least for now).

And the best part? Setting it up is stupidly easy. Wanna see how it works? Let’s get our basic facts straight.

Who Created DeepSeek?

Liang Wenfeng, is a very talented student from Zhejiang University which is located in China. He founded DeepSeek AI. Yes he is the big brain behind Deepseek AI. Before fully focusing on AI, he had co-founded High-Flyer, a hedge fund that used AI for stock trading. Interesting isn’t it? This experience gave him a full understanding of AI beyond finance, how where where AI can be used.

So after learning all this in July 2023, he decided to take things further and established DeepSeek as a dedicated AI company. He thought of developing an advanced AI models can be capable of competing with Google and specially OpenAI. With expertise in both technology and finance, which came to him as a big advantage, he finally created Deepseek that is now taking over the world.

Where Was DeepSeek Founded?

DeepSeek was first started in Hangzhou, Zhejiang, China, a city that is known for being a major in tech and business related areas. (Interesting fact: Alibaba, one of the biggest tech companies in the world, was founded here).

Hangzhou has become a place where all the innovation is done. It has now started attracting AI researchers, tech entrepreneurs, and big investors. Given the circumstances of the place, it was more than obvious that deepseek found a perfect environment to develop its advanced AI models. Being in China also means DeepSeek gets great advantage by getting access to a huge market and specially government support for AI development.

When Was DeepSeek Created?

DeepSeek officially launched in July 2023, but its roots go back way further. If you must know, Liang Wenfeng’s experience with AI at High-Flyer played a big role in shaping the company. It wasn’t an overnight success of course—years of AI research, financial strategy, and technical development led to the very moment when DeepSeek was born. And as of January 2025, their AI chatbot DeepSeek-R1 became the number one app on the U.S. App Store, even surpassing ChatGPT, i mean who would have thought. That’s a pretty insane rise in less than two years!

How Was DeepSeek Built?

DeepSeek didn’t just randomly appear out of nowhere. They actually built it on solid AI tech with a totally different approach. While OpenAI and Google are out here burning billions on massive AI models, DeepSeek was like, “Nah, let’s do it smarter.” These geniuses figured out a way to train AI without even spending a fortune (alot of money saved) but still compete with the big tech industries. Their trick? Smarter training, optimized computing, and actually making AI useful instead of just flashy. Kept costs low, performance high, and boom—no one saw them coming.

main models of deepseek

How to Use DeepSeek AI?

Alright, so you wanna try out DeepSeek AI? Super easy. Just head over to chat.deepseek.com and sign up.

I just use Google to sign in—saves me from remembering another password. (BTW, tons of people are signing up right now, so if it’s a bit slow, don’t stress.)

Once you’re in, you’ll see a simple chat screen. Just type your question and hit enter—that’s it!

For fun, I asked DeepSeek to write a song about squirrels, and in seconds, it gave me a full song. Completely original. No humans involved. Wild, right?

Quick heads-up: Never share personal info with DeepSeek. It learns from chats, and you don’t want your secrets in its brain!

1. Get the App

DeepSeek has apps for iOS and Android. Download them so you can chat with AI anytime. Just be sure you’re getting the real deal, not some scammy knockoff.

2. Use It Like Google (But Smarter)

DeepSeek can search the web for real-time info. Just start a new chat, activate the search mode, and ask about recent events. I asked about a soccer game that happened yesterday, and boom—it gave me the exact score in seconds.

3. Continue the Chat If It Gets Stuck

Sometimes, DeepSeek kinda freezes when searching. If that happens, just hit “Continue” and it’ll keep going until it finds an answer.

4. Read Text from Images

This is actually super useful. Let’s say you have a screenshot of notes or a list of pros & cons about bananas (don’t ask, I just tried it). Instead of typing everything manually, drag and drop the image into DeepSeek, and it’ll turn it into text.

5. Write Code (That Actually Works)

DeepSeek can generate full working code. I asked it to make a snake game, copy-pasted it into an editor, and—guess what? It worked perfectly. You can even test HTML code directly in DeepSeek without leaving the site.

6. Rephrase Text (Because We All Ramble)

Ever written something and thought, “Damn, that’s too wordy”? DeepSeek can simplify your text while keeping all the important details

For example:
Original:
“I go for a walk at this time of day because I don’t want to work all the time and I also like the sunshine, which is rare in winter.”

DeepSeek’s version:
“I take a walk now to enjoy the rare winter sunshine.”.

Short, sweet, and to the point!

7. Write Emails (Without Sounding Awkward)

Not great at professional emails? No worries—DeepSeek can handle that. You just tell it who you’re writing to and what you need, and it spits out a polished email.

Too long? Just say “Make it shorter” and boom—it’s fixed.

8. Solve Homework (Yep, Even Math)

Instead of typing out your question, you can just upload a photo of your homework. And if it’s a tough question (like integrals or logic puzzles), turn on “DeepThink” mode for step-by-step solutions.

(Still double-check the answers though, just in case. )

9. Explain Hard Stuff Like You’re Five

If a topic feels way too complicated, just tell DeepSeek “Explain this to a 5-year-old” and it’ll break it down in simple terms.

I tried it with quantum entanglement, and it actually made sense for once.

10. Be Specific for Better Answers

If you just type “I’m bored, what should I do?”, you’ll get generic responses. But if you say, “Ask me some questions to find the perfect activity”, you’ll get something way more tailored to you.

11. Don’t Stress About Typos

DeepSeek gets typos. No need to backspace like crazy. Just type and move on.

12. Organize Your Chats

You can rename your chats so they don’t all say “New Chat.” Just tap the three dots next to a conversation and give it a proper name.

13. Delete Chats (Individually or All at Once)

If you wanna get rid of old chats, just click “Delete” on a single chat or nuke everything by going into settings and choosing “Delete all chats.”

14. Switch Between Light & Dark Mode

Because dark mode = life (but if you prefer light mode, no judgment). Go to Settings > Theme and pick what works for you.

15. Run DeepSeek PRIVATELY on Your Computer

DeepSeek has a way to run completely offline on your computer. This means:
No internet required
No data sent to DeepSeek’s servers
Full privacy

How to install the DeepSeek R1 model on your device?

if you wanna run DeepSeek AI completely offline, without any of your messages going to the internet, you can totally do that. It’s actually not that hard, but you’ll need to install a few things first.

I’ll walk you through it step by step. Ready?

1: Install Ollama

First things first, you gotta download Ollama. It’s a tool that lets you run AI models on your computer instead of using them online. Just go to ollama.com and install it.

2: Open Ollama

Once it’s installed, open it up. If you’re on Mac, you’ll see a little llama icon pop up, which means it’s running.

3: Install the Command Line Tools

Click the “Install” button to get the command line tools for Ollama. This will let you run AI models straight from your terminal (or command prompt if you’re on Windows).

4: Install DeepSeek AI Locally

Now, time to install DeepSeek itself! You don’t wanna start with a super heavy version right away—your laptop might catch fire (kidding, but you get the point).

So, let’s start small. Open your terminal (on Mac, press Command + Space, type “Terminal,” and hit enter). Then type this command:

ollama run deepseek-r1:1.5b

Hit enter, and boom—it starts installing.

Now, the 1.5B version is kinda like the baby version of DeepSeek. If you want something way more powerful, go for the 32B version instead by running this command:

ollama run deepseek-r1:32b

Takes a bit longer, but it’s worth it if you want better responses.

5: Test It Out

Once everything is installed, you can chat with DeepSeek locally. Just type something like:

ollama run deepseek-r1:32b

Then start talking to it, just like you would online. Try something random like:

“Write a rap about my cat who acts like a dog.”

And see what it comes up with.

6. Turn Off Your Internet (Optional Test)

Wanna see if it’s really running offline? Just turn off your WiFi and try chatting with DeepSeek. It’ll still work because it’s running directly on your device—no internet needed!

7: Stop It When You’re Done

If DeepSeek keeps generating text and won’t stop, just press:

Control + C (on Mac)

And if you wanna completely exit, just type:

/bye

And that’s it! Now you have your own private AI running on your laptop. No internet needed, no data sharing, just you and your AI assistant. Pretty cool, right?

How to Set Up DeepSeek on JanitorAI?

1: Create a DeepSeek API Account

  1. Head over to DeepSeek API and sign up.
  2. Add some credits to your account (around $2–$5 should be enough).

2: Configure JanitorAI

  1. Open JanitorAI and go to API Settings.
  2. Click on Proxy, then select Custom.
  3. Under Model, type: deepseek-chat.
  4. In the Other API/proxy URL field, enter: https://api.deepseek.com/v1/chat/completions?model=deepseek-chat

3: Generate & Apply Your API Key

  1. Go back to DeepSeek API and create a new API Key.
  2. Copy and paste it into the API Key section in JanitorAI.
  3. Click Check API Key—if it turns green, save your settings and refresh the page.

And that’s it! You’re all set to use DeepSeek on JanitorAI.

How to Use DeepSeek via OpenRouter on JanitorAI?

Here’s a quick and easy way to set up DeepSeek on JanitorAI using OpenRouter—and the best part? It’s free!

1: Create an OpenRouter Account

  1. Go to OpenRouter.ai and sign up.
  2. After signing up, go to Settings.
  3. Scroll down to Default Model and select DeepSeek:R1 (Free).

2: Generate an API Key

  1. Go to Keys and create a new API key.
  2. Save it somewhere safe—you won’t be able to see it again!

3: Configure JanitorAI

  1. Open JanitorAI and find a proxy-compatible bot.
    • (To check if a bot is proxy-compatible, see if its character definition is visible. If it’s hidden, it likely won’t work.)
  2. Click on API Settings → Proxy → Custom.
  3. In the Model field, enter: deepseek/deepseek-r1:free (All lowercase, exactly as shown.)
  4. In the URL field, enter: https://openrouter.ai/api/v1/chat/completions (Do not change this or click on anything else.)
  5. Paste the API Key from Step 2 into the API Key section.
  6. (Optional) Add a custom prompt (Jailbreak) if needed.
  7. Click Save Settings.
  8. A pop-up will ask if you want to reset the temperature—click Yes.

4: Adjust Generation Settings

  1. Go to Generation Settings in JanitorAI.
  2. Set Temperature to 1.1.
  3. Set Tokens to 0 (or adjust as needed).

5: Restart JanitorAI & Start Chatting

  1. Close all JanitorAI tabs completely.
  2. Reopen JanitorAI and start chatting!
  3. If it doesn’t work immediately, refresh the page a few times.

And that’s it! You’re now all set to use DeepSeek via OpenRouter on JanitorAI.

How to Install DeepSeek-R1 Locally for Coding & Logical Problem-Solving?

DeepSeek-R1 is a free, open-source AI model built for reasoning, coding, and math-heavy tasks. It’s a solid alternative to OpenAI’s o1 model ($200/month), offering comparable performance while running locally, meaning no cloud fees and complete data privacy.

How to Install DeepSeek-R1 Locally?

1: Install Ollama (MacOS/Linux)

Open your terminal and run:

curl -fsSL https://ollama.com/install.sh | sh
ollama -v  # Check Ollama version

2: Download DeepSeek-R1 via Ollama

Pick the right model based on your hardware:

🔹 7B Model (4.7GB, Good for Consumer GPUs)

ollama run deepseek-r1

🔹 70B Model (Needs 24GB+ VRAM)

ollama run deepseek-r1:70b

🔹 671B Model (Full DeepSeek-R1, Needs 336GB+ VRAM for 4-bit quantization)

ollama run deepseek-r1:671b

3: Set Up Open Web UI (Private Interface)

To access DeepSeek-R1 through a user-friendly interface, install Docker and run:

docker run -d -p 3000:8080 \
  --add-host=host.docker.internal:host-gateway \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

Now, open http://localhost:3000 in your browser and select deepseek-r1:latest.

No cloud tracking, complete privacy!

Testing DeepSeek-R1 Locally: Try generating a Snake game in Python. DeepSeek-R1 creates efficient code, though it takes about 3 minutes to respond. This slight delay is worth it for better logical reasoning and step-by-step explanations.

How to Integrate DeepSeek-R1 Into Your Projects?

1. Local Deployment (Privacy-First, No API Key Needed)

Run DeepSeek-R1 as an OpenAI-compatible endpoint using Ollama:

import openai

# Connect to local Ollama instance
client = openai.Client(
    base_url="http://localhost:11434/v1",
    api_key="ollama"  # No authentication required
)

response = client.chat.completions.create(
    model="deepseek-r1:XXb",  # Replace "XX" with your model size (7b, 70b, etc.)
    messages=[{"role": "user", "content": "Explain blockchain security"}],
    temperature=0.7  # Adjust creativity level
)

2. Cloud API (For Scalable, Production Use)

If you need cloud-based access, use DeepSeek’s official API:

import openai
from dotenv import load_dotenv
import os

load_dotenv()
client = openai.OpenAI(
    base_url="https://api.deepseek.com/v1",
    api_key=os.getenv("DEEPSEEK_API_KEY")
)

response = client.chat.completions.create(
    model="deepseek-reasoner",
    messages=[{"role": "user", "content": "Write web scraping code with error handling"}],
    max_tokens=1000  # Limit costs for long responses
)

Points to be noted:

  1. Don’t pressure bot creators to make their bots proxy-compatible—it’s their choice.
  2. Quick way to check: If a bot’s character definition is visible, it’s likely proxy-compatible. If hidden, it likely isn’t.
  3. DeepSeek logs messages (both yours and the bot’s) to improve the AI model.
  4. The data is anonymous, but remember—nothing online is 100% private.
  5. Want to change your IP address? Unplug your router for five minutes, then reconnect.
  6. As long as you’re not sharing personal/sensitive info, you’re fine. (No, your AI crush won’t leak your secrets.)
  7. To use a paid DeepSeek model:
    • In Step 2, select a paid model instead of the free one.
    • Skip Step 6—the API key will automatically detect your selected model.
  8. DeepSeek has a daily token limit. If you hit it, just wait until the next day.
  9. Some paid APIs bypass this, but most users find the free version works just fine.
  10. To remove DeepSeek’s “Thinking” Delay
    • Go to OpenRouter Settings → Ignored Providers.
    • Add Targon and save.

How To Run DeepSeek In R?

Alright, so if you ever wanna run DeepSeek in R, here’s the deal—it’s actually pretty simple once you get the hang of it. Let me walk you through it step by step.

1: Grab Your API Key

First things first, you gotta sign up on DeepSeek’s platform (just head over to platform.deepseek.com). Use your email to make an account. Once you’re in, top up your balance—just a few bucks will do. If your card isn’t working because of country restrictions, PayPal is your best bet (you can still pay without making an account).

Now, go to API Keys in the menu and create a new one. It’ll look something like this: sk-xxxxxxxxxxxxx. Copy that somewhere safe—you won’t be able to see it again.

2: Install Some R Packages

Before you do anything in R, install these two packages (just copy-paste these into your R console):

install.packages("httr")
install.packages("jsonlite")

These will let you send requests to the DeepSeek API and handle the responses.

3: Set Up DeepSeek in R

Now comes the fun part—actually making DeepSeek respond to your prompts! You’ll need your API key and a question you want to ask. Here’s a basic example:

library(httr)
library(jsonlite)

apiKey <- "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxx"
prompt <- "How to summarize data using dplyr."
model <- "deepseek-chat"

response <- POST(
  url = "https://api.deepseek.com/chat/completions",
  add_headers(Authorization = paste("Bearer", apiKey)),
  content_type_json(),
  encode = "json",
  body = list(
    model = model,
    temperature = 1,
    messages = list(list(
      role = "user",
      content = prompt
    )),
    stream = FALSE
  )
)

cat(content(response)$choices[[1]]$message$content)

Boom! That should get DeepSeek talking. For example, if you ask about summarizing data in dplyr, it might respond with something like this:

summary_df <- df %>%
  summarize(
    mean_value = mean(value),
    median_value = median(value),
    max_value = max(value),
    min_value = min(value)
  )

4: Make It Even Better with a Function

Typing all that code every time is a pain, so let’s make a function that does everything for us:

deepseek <- function(prompt,
                    modelName = "deepseek-chat",
                    temperature = 1,
                    apiKey = Sys.getenv("deepseek_API_KEY")) {
  
  if(nchar(apiKey) < 1) {
    apiKey <- readline("Paste your API key here: ")
    Sys.setenv(deepseek_API_KEY = apiKey)
  }
  
  response <- POST(
    url = "https://api.deepseek.com/chat/completions",
    add_headers(Authorization = paste("Bearer", apiKey)),
    content_type_json(),
    encode = "json",
    body = list(
      model = modelName,
      temperature = temperature,
      messages = list(list(
        role = "user",
        content = prompt
      ))
    )
  )
  
  if(status_code(response) > 200) {
    stop(content(response))
  }
  
  trimws(content(response)$choices[[1]]$message$content)
}

Now, all you gotta do is call it like this:

cat(deepseek("square of 29 plus 67"))

and DeepSeek will handle the rest.

5: Keep the Conversation Going

By default, DeepSeek forgets everything you said earlier, which kinda sucks. But we can fix that by keeping a chat history:

chatHistory <- list()

deepseek <- function(prompt) {
  chatHistory <<- append(chatHistory, list(list(role = "user", content = prompt)))

  response <- POST(
    url = "https://api.deepseek.com/chat/completions",
    add_headers(Authorization = paste("Bearer", apiKey)),
    content_type_json(),
    encode = "json",
    body = toJSON(list(messages = chatHistory), auto_unbox = TRUE)
  )

  answer <- trimws(content(response)$choices[[1]]$message$content)
  chatHistory <<- append(chatHistory, list(list(role = "assistant", content = answer)))

  return(answer)
}

Now, check this out:

cat(deepseek("2+2")) # It’ll return: 4
cat(deepseek("square of it")) # Now it remembers and returns: 16
cat(deepseek("add 3 to result")) # And it returns: 19

Cool, right? You’ve basically made DeepSeek act like a proper chatbot inside R! So yeah, that’s how you can run DeepSeek in R without breaking a sweat. Let me know if you need help setting it up!

DeepSeek vs ChatGPT: Which One’s Better?

Both are top AI chatbots, but they shine in different ways:

  • DeepSeek is faster, cheaper, and great for technical work like coding and data analysis.
  • ChatGPT is more versatile, creative, and super user-friendly, making it perfect for writing, brainstorming, and casual chats.
  • DeepSeek: Just $0.003. No joke.
  • GPT-4o: Around $1 for 10 responses (~1,000 tokens).

Basically, it depends on what you need!

Which One Should You Go For?

DeepSeek is your best bet if you need a fast and budget-friendly AI for coding, data analysis, or other technical tasks.

ChatGPT is the way to go if you’re looking for a well-rounded AI that’s great at creative writing, brainstorming, research, and just casual conversation.

Pretty much depends on what you need!

Did DeepSeek Train on ChatGPT’s Work?

I dont know if you have heard this but there’s this big rumor going around that DeepSeek might have learned from ChatGPT’s responses instead of training on its own original data. This process is called “distillation” an AI learning from another AI instead of real-world sources. In simple words a cheap knockoff.

If true, DeepSeek isn’t really standing on its own, it’s just borrowing from OpenAI’s efforts. It is unethical but who cares? This could lead to serious legal issues if OpenAI steps up.

DeepSeek hasn’t come out and confirmed (or denied) these claims yet, which just makes people even more suspicious.

Censorship & Political Influence

As we know that DeepSeek AI is developed in China, it is said that its responses are reportedly filtered to align with Chinese government regulations. In short we might now be getting the original version of any information that we search on Deepseek. Rather we are getting a Chinese version of it that aligns with their government regulations.

To clarify this information i went on reddit and scrolled through thousands of threads. Users have noticed that DeepSeek avoids or refuses to discuss politically sensitive topics like Tiananmen Square, Taiwan, or human rights issues.

And obviously this would raise concerns among users as to why AI should control the information that we consume. Another section of people believe that AI should not have its own opinions that reflects on government topics hence not providing sensitive topics information. No doubt ChatGPT also have some content restrictions but DeepSeek’s censorship appears to be more extensive.

This sparks a debate—should AI remain neutral, or should it follow government-imposed rules? If DeepSeek expands globally, will it enforce the same level of censorship everywhere?

Privacy & Security Concerns

Okay, so here’s the deal—DeepSeek AI has been caught up in some privacy drama. People are saying that some of its backend tech might be connected to ByteDance (yep, the same company behind TikTok), and you know how that usually goes.

Some researchers think DeepSeek could be collecting user data and, worst-case scenario, sharing it with Chinese authorities. There’s also this whole thing about AI being used for surveillance—like, imagine your chats being monitored without you even knowing. That’s sounds creepy. Even Microsoft, which hosts DeepSeek on Azure, is keeping a close eye on how it’s being used.

Well, remember how some countries banned TikTok over data privacy concerns? Now, people are asking the same questions about DeepSeek—like, can we really trust it with our data?

DeepSeek’s Availability by Country

Keeping up with the latest news, DeepSeek AI has been banned or restricted in multiple countries and government agencies over concerns about data privacy, security risks, and possible data sharing with Chinese authorities.

Where DeepSeek AI Is Banned or Restricted?

  • Italy – Italy blocked DeepSeek AI because they didn’t get enough clarity on how personal data is used. Way to go Italy, having their priorities in line for people.
  • South Korea – Government employees can’t use DeepSeek’s AI chatbot due to security concerns. So this means not banned just restricted.
  • Australia – Just like South Korea, Aussie government agencies have also restricted access for employees over security worries.
  • United States (New York State) – New York has banned the DeepSeek AI app from all government devices, mainly due to data privacy and censorship concerns.
  • India – The Ministry of Finance has straight-up prohibited government employees from using DeepSeek AI, citing national security risks.
  • Taiwan – DeepSeek AI is banned here too, over concerns that data might leak to foreign entities.

S o many countries restricting due to security reason, there is definitely something fishy going on. Since this list could grow, keeping an eye on official updates is a good idea.

Reasons for Bans and Restrictions

Data Privacy – People are worried that DeepSeek AI might be collecting user data, which can be accessed by Chinese authorities. It all because of some China’s data-sharing laws. I admit it is a level of concern, invading privacy using data, we are safe nowhere.

Security Risks – Also, there’s concern that the AI might have vulnerabilities that could expose sensitive info, especially in government sectors. I mean we don’t have any control over anything so how can we trust AI from not exposing our secrets.

Censorship & Content Control – Some have noticed that DeepSeek AI’s responses seem to align with Chinese government views. This is something that we need to get concerned about because this would mean that there is potential information manipulation. We don’t get original raw content instead some Chinese version that they want us to read.