China ’s DeepSeek R1 and the USA’sOpenAI o1are both reasoning simulation . Instead of answering questions immediately , they take meter to think through the prompt using their logical thinking operation , run to better and more accurate answers . These good example are generally effective at manage complex questions come to to coding , math , science , or anything require serious reasoning skills .
Until now , OpenAI ’s o1 model has been leading the diligence in reasoning capability . However , it is a shut - source AI example approachable only through a$20 give subscription . Google is also turn on its own logical thinking simulation called 2.0 Flash Thinking , but it is still in genus Beta . While promising , it has n’t quite reached the grade of o1 and is only available through Google AI Studio . We will put it through the paces when it is available .
On the other side of the function , DeepSeek from China has release its R1 model this hebdomad , which is mostly on par with OpenAI ’s o1 model but surpass it in some other area . It has become the public lecture of the town ever since . Unlike OpenAI ’s o1 model , R1 is open - source , free to apply , and has achieved o1 bench mark at just3 % of the cost . Not surprising since China has always been good at doing affair in a very monetary value - effective way . Even the developerAPIs are 90%-95 % cheapercompared to the o1 manakin .
OpenAI o1
But how safe is the R1 AI model and can it really beat the o1 model by ChatGPT ? permit ’s ascertain out using a twain of prompt .
DeepSeek R1 vs OpenAI o1
To test the claims , we valuate both OpenAI ’s o1 model and DeepSeek ’s R1 poser with various prompts involve strong reasoning acquisition to see if DeepSeek has truly delivered o1 - layer performance or even stand out it .
1. Puzzle-Based Reasoning
I started the comparing with a definitive puzzler - style question that does not even have a working resolution .
So let ’s see which model can figure out that it does not have an solution . While theo1 model exact just 16 secondsto think , DeepSeek take 120 mo . However , both exemplar came to the right conclusion , saying there is no elbow room to visualize out who is a horse and who is a knave . I found DeepSeek ’s account much easier to sympathise than o1 ’s confusing narrative .
The good part about DeepSeek is you cansee its integral logical thinking process , which is quite compelling . It reason through like we humanity do and tries to solve the payoff in various means multiple time . The summons is written from DeepSeek ’s perspective lead in a much better and fascinating user experience . For example , here ’s part of the textual matter fromDeepSeek ’s thought process :
OpenAI o1
Interesting , right ?
Verdict : Both the AI models got the result right on . While ChatGPT ’s o1 is fast , DeepSeek ’s R1 is more thorough and provides a elementary account that humans can understand and digest more promptly .
2. Math Problem
Next , I have a toilsome math - related doubtfulness that can take at least 30 - 50 whole tone to find oneself the resolution .
Both simulation predicted the answers right . However , DeepSeek provide an precise answer , mentioning 3.18 years , whereasChatGPT rounded it offto 3.2 twelvemonth . But o1 was much faster , thinking for just 5 indorsement , whereas DeepSeek pick out 53 bit to go far at the answer .
Verdict : Both the models again put up the correct answer , however , o1 Model is much faster . On the other DeepSeek shares the total calculation and the exact solution which can make all the deviation when it come to math , science , and mystifying space .
DeepSeek R1
3. Solving a Sudoku Puzzle
Who does n’t love a sudoku puzzle ? For the third interrogative sentence , I uploaded a Sudoku puzzle as an range from the r / sudoku subreddit to both the AI models postulate them to solve it .
clear a Sudoku puzzle seems too much for any AI reasoning modeling . However , if the mannequin have codification instruction execution capabilities , they can generate or use an subsist computer code in their database and execute it to work the mystifier . For exercise , Gemini 1.5 Pro can clear Sudoku teaser . However , both ChatGPT o1 and DeepSeek R1 example endeavor to solve the Sudoku with just reasoning , and here are the results .
DeepSeek reasoned and took 68 second before saying thegrid was not perfect , even though it was . I uploaded two other Sudoku puzzle , and the effect were the same . This is probable because DeepSeek ’s visual sensation capabilities are subpar . While it can argue through problems , it struggles to interpret uploaded images .
OpenAI o1
OpenAI , on the other hand , thought for more than 5 minutes and provide awrong result . I upload two other Sudoku puzzler just like on DeepSeek . However , once , it did pull off to give the correct answer in 5 seconds , indicating that thesolution was already in its breeding information .
At least o1 simulation was able-bodied to read the double and uploaded files better than DeepSeek R1 , however , both models could n’t solve any sudoku puzzlecorrectly .
Finally , I enrol the sudoku puzzle in the text data formatting , with no range . OpenAI again institute the solution useable in its grooming datum , whereas DeepSeek went through the logical thinking process take 280 seconds and again came up with the wrong answer . So we can reason out it ’s not just image potentiality , Sudoku puzzler are unsoluble for the current hatful of AI reasoning example .
DeepSeek R1
Verdict : Both theoretical account give out to arrive at an answer through reason .
Also Read :
4. Creating a Flowchart
I take both AI reasoning manakin to make a flowchart of how the OpenAI ’s Operator do work . This can be an event for the o1 model as itcannot enter the internetandOperatoris a late development not available in its training data . However , DeepSeek ’s reasoning model can get at the internetso allow ’s see what it can do .
As expect , o1 created ageneric flowchartof how OpenAI ’s LLM fashion model work , not the Operator model . The flowchart was also puzzling and barebones . DeepSeek searched online for information about the Operator and generated a flow sheet as requested .
Verdict : DeepSeek R1 wins by a landslide .
OpenAI o1
5. Programming Task
To labialize off our DeepSeek R1 vs OpenAI o1 comparison , I go for a programming - related query this time .
It ’s a simple-minded challenge that can be easily completed with existing modules . OpenAI o1 fashion model used thetransformers pipelinemodule and sharedhow to install that moduleon my PC before running the code . Whereas DeepSeek ’s R1 directlyprovided the code with no stepsand used avaderSentiment modulewhich I had never used .
After installing both modules and running the code , we could separate DeepSeek ’s implementationfollowed the instructionsbetter . For case , the app produce by o1 did not provide a proper explanation for its sentiment categorization , whileDeepSeek ’s app gave well-defined cause . Additionally , DeepSeek ’s appworked in real - metre , analyse the input as you typecast , whereaso1 want click the Analyze button .
DeepSeek R1
However , neither model could understand the sarcasm ! But for the most part , they sustain the business done .
Verdict : DeepSeek R1 for follow the instruction accurately .
Final Verdict: ChatGPT o1 vs DeepSeek R1
As you may see , the only query DeepSeek fail to reply correctly was the Sudoku puzzle , which OpenAI also go bad at . Except for that , DeepSeek ’s R1 model consistently cater easier - to - understand account and exact answer following instructions to the T. All while transparently showcasing its abstract thought outgrowth . On top of that , it ’s free to use and open - root pee-pee it accessible for all .
We have also try out both logical thinking models in twenty-four hours - to - day usage , and DeepSeek is on par with OpenAI ’s o1 model , often surpassing the latter ’s pay plans .
DeepSeek ’s claims hold true and users can confidently rely on it as a alternate for the o1 theoretical account . However , OpenAI also has an o1 Pro model which costs $ 200,and is preparing tolaunch the o3 modelsoon , so the narrative may lurch soon enough . But for now , considering the Mary Leontyne Price , open - source availability , and public presentation , we can reason : DeepSeek R1 > OpenAI o1 .
OpenAI o1
All Secret And Hidden Commands for Alexa (May 2025)
How to Turn ChatGPT AI Action Figure Into a Video…
7 AI Tools That Research for You: Stop Drowning in…
How to Create ChatGPT Action Figures and Turn Yourself Into…
4 Ways to Use ChatGPT to Generate YouTube Thumbnails –…
ChatGPT Not Generating Ghibli Images? Here’s What to Do
We Put ChatGPT and Gemini’s Image Generators Head-to-Head – Here’s…
How to Turn Ghibli Images Into Videos Using AI for…
We Tested Tens of AI YouTube Thumbnail Makers to Discover…
I Created My Own Apps Using AI — Here’s What…
DeepSeek R1
OpenAI o1
DeepSeek R1
