DeepSeek is the brand new AI chatbot on all people’s lips and is at present sitting on the prime of Apple’s App Store within the US and the UK. A totally free AI mannequin constructed by a Chinese start-up, DeepSeek needs to make AI much more accessible to the plenty by providing a competitor to OpenAI’s ChatGPT o1 reasoning model and not using a payment.
New AI apps seem on the App Store virtually every day, and there’s typically hype round a brand new mannequin launch as folks search for the following different to ChatGPT. Whether you’re an avid person of OpenAI’s software program otherwise you desire to make use of Google Gemini, there’s an AI device for everybody, and DeepSeek needs to be the following icon on your house display screen.
After seeing DeepSeek throughout my newsfeed, I knew I needed to give the brand-new AI a go and see if it was nearly as good as individuals who made it out to be on-line. I pitted DeepSeek V3 and DeepThink R1 towards ChatGPT 4o and o1 to see simply how good the brand new king of the App Store actually is.
AI life hacks
In this take a look at, I wished to get a full really feel for all the things DeepThink presents in comparison with ChatGPT, so I solely thought it was honest to make use of the AI chatbot the identical method I’d use AI in my every day life.
Recently, I’ve been desirous to get assist from AI to create a every day schedule that matches my wants as an individual who works from house and must take care of a canine. Up till just lately, my accomplice additionally labored from house, so it was a lot simpler to separate duties, however she is now working from an workplace, and I have to discover a strategy to juggle life, work, and my French Bulldog Kermit.
I requested ChatGPT o4 and DeepSeek V3 to create a every day schedule with some info on after I get up, my canine’s potty routine, and a quick breakdown of my workflow. Both created wonderful schedules that I might genuinely see myself utilizing every day. However, ChatGPT’s reminiscence function made OpenAI’s schedule much more coherent.
I had beforehand informed ChatGPT that I wish to overview AI information and developments at 9 am, and 4o applied that info from a earlier chat into my morning routine. DeepSeek, alternatively, can solely keep in mind info from the identical chat and couldn’t deliver again info from earlier chats to assist with its reply.
(ELI5) Explain like I’m 5
Next, I wished to ask each AI chatbots in regards to the NFL Playoffs, contemplating we now know the 2 groups that can face one another at Super Bowl LIX. I requested DeepSeek and ChatGPT to present me a 200-word rundown of the NFL playoffs and the way it works. Both supplied wonderful info that gave me a full understanding of how the seeding system works and the journey a staff must take to make it to the Super Bowl.
ChatGPT opted for a 200-word paragraph, whereas DeepSeek broke info down into bullet factors. I did discover that ChatGPT gave me extra context on how groups turn into a Wild Card, however the distinction between the outcomes is pretty minimal and also you’ll like one higher than the opposite purely based mostly on private choice.
Problem solver
Now that we’ve coated some easy AI prompts, it’s time to get all the way down to the nitty gritty and check out DeepThink R1, the AI mannequin that has everybody speaking. People on-line are saying DeepSeek’s free reasoning mannequin is nearly as good as ChatGPT’s o1, which is free in small doses however requires a paid subscription to entry commonly.
To take a look at the AI chatbots’ reasoning capabilities, I regarded for among the hardest issues I might discover, and I’m shocked by among the outcomes:
Question 1: Find the lacking phrase: Apple, Red, Coal
This isn’t a very exhausting query, particularly contemplating the supply materials was multiple-choice with totally different shade choices. I opted to keep away from giving R1 and o1 multiple-choice solutions and as an alternative simply wrote the query and hit enter.
ChatGPT o1 took 1 minute and 29 seconds to find out the reply, and it discovered hyperlinks between the phrases and the fairytale Snow White. The mannequin determined to reply based mostly on this quote, “her lips were red as blood, her hair was black as coal, and her skin was white as snow.” Based on this quote o1 selected Snow because the lacking phrase reply. While its thought course of was intelligent, it wasn’t the reply I used to be in search of.
DeepThink R1, alternatively, took 1 minute and 14 seconds to reply, and it managed to guess the suitable phrase: Black. Apple is purple; coal is black. Impressive, to say the least.
Question 2: 1. Finish the sequence: 1, 2, 4, 8, ? 2. Finish the sequence: home, Saturn, canine, burger, ?
These two sequences are fully unrelated, however I believed it could be attention-grabbing to ask back-to-back inquiries to see what occurs. While the primary sequence could be very simple, the second is not possible (they’re simply three random phrases). Would ChatGPT o1 or DeepThink R1 have the ability to discover the lure?
Well, no. Both reasoning fashions tried to seek out a solution and gave me a very totally different one. DeepThink R1 answered “yellow” as a result of it thought the phrases had been associated to their shade (white home, yellow Saturn, brown canine, yellow burger). ChatGPT o1, alternatively, answered “car” as a result of it discovered the sequence virtually not possible however determined to supply solutions based mostly on “a common puzzle approach.” The strategy it selected to supply up was linking every merchandise into the larger class it belongs to (home = constructing, Saturn = planet, canine = animal, burger = meals, and automobile = automobile).
Ultimately, each reasoning fashions had been improper, and neither responded by saying there have been too many variables to present an correct reply.
Question 3: Hummingbirds inside Apodiformes uniquely have a bilaterally paired oval bone, a sesamoid embedded within the caudolateral portion of the expanded, cruciate aponeurosis of insertion of m. depressor caudae. How many paired tendons are supported by this sesamoid bone? Answer with a quantity.
For the ultimate query, I made a decision to ask ChatGPT o1 and DeepThink R1 a query from Humanity’s Last Exam, the toughest AI benchmark on the market. To a mere mortal like myself with no data of hummingbird anatomy, this query is genuinely not possible; these reasoning fashions, nevertheless, appear to be up for the problem.
O1 answered 4, whereas DeepThink R1 answered two. Unfortunately, the right reply isn’t out there on-line to forestall AI chatbots from scraping the web to seek out the right response. That mentioned, from some analysis, I imagine DeepThink could be proper right here, whereas o1 is simply off the mark.
DeepSeek vs ChatGPT?
So, I’ve run a number of prompts and used each chatbots for an intensive period of time, however what’s the higher choice? According to the solutions I acquired from prompts, DeepThink R1 is a wonderful free reasoning mannequin that makes you query whether or not it’s value paying to entry o1 commonly. DeepSeek is just out there on the net, iOS App Store, and Play Store, so if you wish to use a standalone Mac app or iPad app, you’ll want to attend for the corporate to launch one.
According to Humanity’s Last Exam, DeepThink R1 outperforms ChatGPT o1 with a 9.4% accuracy price in comparison with OpenAI’s 9.1%; it’s a marginal distinction, however contemplating one is totally free, it could sway you in the direction of utilizing the brand new child on the block.
Personally, I’ll be sticking with ChatGPT as a result of I don’t have huge necessities for reasoning fashions, and I rely closely on the reminiscences function, which permits the AI chatbot to reference earlier conversations. I additionally like the truth that ChatGPT has a standalone Mac and iPad app, in addition to the power to generate photographs with among the finest AI picture mills, DALL-E.
DeepSeek is only text-based and lacks multi-modal capabilities, however contemplating how new it’s, that is an extremely promising begin to a real challenger for OpenAI’s AI crown.