Meta AI vs Gemini vs ChatGPT: The Ultimate Comparison

Reading time icon 45 min. read


Readers help support MSpoweruser. We may get a commission if you buy through our links. Tooltip Icon

Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more

Meta AI vs Gemini vs ChatGPT

Comparing Meta AI vs Gemini vs ChatGPT is difficult.

But not impossible.

They’re all multi-purpose AI assistants with extensive capabilities from the world’s top companies. Meta, Google, and OpenAI (in partnership with Microsoft) invest billions to advance these platforms.

That’s precisely why I decided to make this comparison. Let’s see who is more productive in investing (almost) unlimited resources and what it means for the end users.

Meta AI vs Gemini vs ChatGPT

Introduction Meta AI vs Gemini vs ChatGPT

Launched in April 2024, Meta AI is the latest AI chatbot (coming from top tech companies) available for public use. It can generate text and images and is accessible through all Meta social media platforms (Facebook, Instagram, WhatsApp).

Introduced to the public in December 2023, Google’s Gemini has since been optimized into three versions. The Ultra model is used for complex tasks, the Pro version is an adaptable assistant for various tasks, and Nano is used for on-device efficiency.

The last weeks of 2022 marked the public release of ChatGPT and, with it, a new era for technology. It was the first platform to reach 1 million users within 5 days and the user count has only grown since then.

Now let’s see how they all compare in what matters the most: the quality of output in various use cases.

Please note: I used the most advanced models for each competitor (publically available at the moment of writing). These are Gemini Advanced (1.0 Ultra), ChatGPT Plus (GPT-4), and Meta AI (Llama 3).

Content Creation

When I hear generative AI, the first thing that comes to mind is content creation, particularly textual content. Let’s see how helpful these AI models will be in writing a short piece on the topic of my choice.

Prompt: I need a short (150 to 200 words) text about camping. I want you to make it cheerful, engaging, and ready to use on my camping website. Under any circumstances, don't use the following words: 'dangerous, wild animals, cold weather' and use words that will create good emotions in the reader.

Meta AI output:

Get Ready for an Unforgettable Camping Adventure!

Imagine waking up to the sweet songs of birds, surrounded by lush greenery and the warmth of a golden sunrise. Our campsite offers the perfect escape from the hustle and bustle of city life, where you can reconnect with nature and create lifelong memories with friends and family.

Enjoy a tranquil retreat under the starry night sky, with the soothing sounds of a crackling campfire and the aroma of s’mores wafting through the air. Our scenic trails invite you to explore and discover the beauty of the great outdoors, with breathtaking views and hidden gems waiting to be uncovered.

Whether you’re a seasoned camper or just starting out, our campsite has everything you need for an adventure-filled getaway. So pack your bags, grab your gear, and get ready to make some unforgettable memories in the great outdoors!


Gemini output:

Crackling campfire, starry nights, and the sweet smell of pine – escape to an unforgettable adventure under the wide-open sky! Camping is your chance to reconnect with nature, breathe in the fresh air, and create memories that will last a lifetime. Imagine toasting marshmallows under a blanket of twinkling stars, listening to the symphony of crickets chirping, and waking up to a breathtaking sunrise. Whether you’re seeking a romantic getaway, a fun-filled family trip, or a solo escape to recharge, there’s a perfect campsite waiting to be your home away from home. So pack your sense of wonder, grab your camping gear, and get ready to experience nature’s beauty up close!


ChatGPT output:

Camping is your ticket to a world of serene beauty, where sparkling lakes meet expansive skies. It’s an adventure that invites you to unplug from the digital world and reconnect with nature. Imagine waking up to the gentle sounds of a nearby stream and the sweet chirping of birds, as you breathe in the fresh, crisp air.

During the day, explore the scenic trails or partake in a peaceful picnic under the shade of towering trees. As the sun sets, gather around a warm campfire, sharing stories and roasting marshmallows with friends and family. The night sky, adorned with stars, offers a spectacular view, perfect for stargazing.

Camping not only rejuvenates the spirit but also strengthens bonds with loved ones, as you create cherished memories together in the great outdoors. It’s a joyful escape to simplicity and happiness, where every moment is an opportunity to smile, laugh, and enjoy the beauty around you. So pack your bags, grab your tent, and set off on a delightful journey into nature’s embrace!


All outputs are cheerful and engaging, including a call to action (pack your bags, etc.), making them ready for a website.

The only AI tool that respected my desired word count was ChatGPT.

None of the models added the words I told them not to, so the last parameter is emotion.

Although subjective, Gemini was the only model that focused on experience rather than plain and generic information.

The output was relatively short, but the point nowadays isn’t on quantity but quality. That’s why Gemini did the best job in text content creation.

Image Generation

According to the number of Google searches, image generation is the most popular use case for AI models.

Although Meta AI, Gemini, and ChatGPT support image generation, they are primarily chatbots. This means that the output quality will depend on the quality of the context built between the user and AI.

This time, the prompt is: Generate an image that visualizes the scenario you just described.

The prompt was sent to the chat I used for the text generation task above.

Meta AI output:

Meta AI image generation

Gemini output:

Gemini Image Generation

ChatGPT output:

ChatGPT image generation

Please note that I didn’t add any information about the style. The only input was my prompt on content creation and the AI’s answer.

ChatGPT gave me an illustration with some apparent mistakes (have a closer look at the characters’ hands), Gemini created an illustration (suitable for 5th-grade art class), and Meta AI produced something I’d consider using on a website.

Let’s try this one more time to eliminate any doubts!

This time, I’ll use the following prompt in a new chat:

Generate an image of a camping site suitable for a website about camping. Focus on details that make the viewer feel good and safe.

Meta AI output:

Meta AI image generation only prompt

Gemini output:

Gemini image generation only prompt

ChatGPT output:

ChatGPT image generation only prompt

ChatGPT did a much better job this time but left a few inconsistencies (e.g., a very small tent on the right, weirdly shaped bench and table). Compared to the others, Gemini wasn’t up to the task once again, as the image looked like a quickly-made collage.

Meta AI generated the best image in both cases (contextual chat history and prompt-only). The end result is still not perfect, but there’s potential with a bit more work.

Logo Generation

Let’s see how these three AI models work when creating a brand logo for my imaginary camping site. First, I’ll do it contextually (in the same chat I made the text and image) with the following prompt:

Generate a logo for my camping site called Camper Heaven, combining all the details we discussed.

Meta AI output:

Meta AI logo generaiton only prompt

Gemini output:

Gemini logo generation

ChatGPT output:

ChatGPT logo generation

Since I provided the campsite’s name, I’ll focus only on outputs that include it. (Many generators struggled with text in logo generation until recently.) All of the competitors managed to create an output with the correct text!

Let’s try it one more time in a new chat with the following prompt:

Generate a logo for a camping site called Camper Heaven. Make it simple, using a maximum of three colours, and without too many details. It should evoke good emotions and represent safety.

Meta AI output:

Meta AI image generation only prompt

Gemini output:

Gemini logo generation only prompt

ChatGPT output:

ChatGPT logo generation only prompt

This time, Meta AI failed to write the text accurately and didn’t reduce the number of details. Gemini did well by making a simple logo and matching its style with the font. ChatGPT went for an ultra-simple version and chose blue as the primary color.

If I had to choose one of these three, Gemini would be my choice for logo creation. It has correct text, fewer details (when asked to do so), and a single-color background.

Coding and Debugging

To test coding and debugging capabilities, I’ll first create a simple Python project, run it, and see how it goes after.

Prompt: Generate a Python code that asks the user to input the dimensions of a room in meters (x and y) and calculate the surface based on that. After that is completed, the system should ask the user if they want to add a room height in meters (z) and, based on that, calculate the volume of the space. Finally, it should ask users if they want to know how many cubes 2 by 2 meters can they fit in. If they respond positively, the answer should be given. I will use an online compiler to test this code.

I’ve intentionally omitted the information about the third dimension of the cube.

Meta AI code result:

Python code sample Meta AI

Gemini code result:

Python code sample Gemini

ChatGPT code result:

Python code sample ChatGPT

Meta AI created code and didn’t comment on the third dimension of the cube, while Gemini and ChatGPT did. The code generated by Gemini immediately shows the number of cubes without asking the user if they want to see it. So far, ChatGPT has performed the best.

Since none of the models added data validation (causing an error if I input ‘five’ instead of ‘5’), I sent the following prompt:

Modify this code so it contains data validation. Make it fool-proof so the system never gives an error and always works flawlessly.

It’s a bit of a bold request from my side, but let’s see what happens.

All of the codes now contained the data validation. However, Gemini didn’t add the explanation for the user, so instead of seeing something like ‘Please enter a valid number,’ I got the ‘could not convert string to float‘ error instead.

At this point, it’s safe to say that ChatGPT did the best job with coding and debugging.

If you want to see the complete final code, check:

Please note: Although this was a trivial example, it was intended to illustrate how closely the instructions were followed. Codes that included extra steps (e.g., error handling) earned additional points.

Web Design

The next task I’ll put the competing AI models against is web design. Based on the prompt below, I’ll use an online compiler to evaluate the HTML and CSS that they generate.

We’ll start with the contextual chat:

Generate a web design (HTML + CSS) of a camping website based on the details you already know.

Meta AI output:

Meta AI web design contextual

Gemini output:

Gemini web design contextual

ChatGPT output:

ChatGPT web design contextual

I’ll start with the mistake that first caught my attention. Meta AI put the year 2023 in the footer, which is a big red flag (but not from the web design point of view.) I even had to repeat the prompt because it initially generated an image instead of the code.

Gemini generated a simple output but didn’t contrast the H1 and the background nicely. The model noted that the logo should be added manually, which is a good reminder if you’re a beginner.

So far, I liked the ChatGPT output the best. Gallery with the note to add photos (even a hint for what to add), nice colors, and a layout for a contact form.

For better insight, let’s create a landing page for summer camping in 2024 with the following prompt in a new chat:

Generate a landing page design, HTML and CSS, for a campsite called Camper Heaven. It's for the summer of 2024 and should include the following photos: Two people in front of the tent at night (link), a pot above the fire close-up (link), a premium house in the forest with a big fire in front (link). I leave everything up to you, but be creative and make excellent styling effects. Also, focus on good CTAs.

Meta AI output:

Meta AI noncontextual web design

Gemini output:

Gemini noncontextual web design

ChatGPT output:

ChatGPT noncontextual web design

Meta AI didn’t manage to provide a design with images. Additionally, the layout isn’t adjusted accordingly, and the contrast between the title and the background is nonexistent.

Gemini did a fair job but again failed to create contrast. This time, it’s the top menu in question, and since it wasn’t styled, the default blue was left hanging above the green background.

Although the noncontextual design lacks appropriate colors, ChatGPT produced the most accurate design. It included the CTA and added contact information in the footer (not visible in the screenshot).

Recipe Generation

One can quickly get hungry after creating many copies, photos, and websites. Let’s see what ideas we can get from our three AI models.

I have the following ingredients: 1 liter of milk, three eggs, 500 grams of flour, 250 grams of salami, and a coupon for three additional ingredients (it can be whatever I decide in whatever quantity). Suggest a meal, a visualization (generate an image), and a recipe to follow to create it.

Meta AI dish:

Creamy Salami Quiche

Meta AI recipe visualization

Recipe:

  1. Preheat your oven to 375°F (190°C).
  2. Roll out the flour to make a pastry crust and place it in a 9-inch tart pan.
  3. In a large bowl, whisk together the milk, eggs, and a pinch of salt.
  4. Add the grated cheese, chopped parsley, and melted butter to the bowl and mix well.
  5. Slice the salami into thin strips and add it to the bowl.
  6. Pour the mixture into the pastry crust.
  7. Bake for 35-40 minutes or until the quiche is set and golden brown.
  8. Serve warm and enjoy!

Gemini dish:

Savory Salami & Herb Crepes with Creamy Cheese Sauce

Gemini recipe visualization

Recipe:

Make the Crepes:

  • Whisk together the milk, eggs, flour, and salt until smooth.
  • Heat a lightly greased pan over medium heat.
  • Pour a thin layer of batter and swirl to coat the bottom of the pan.
  • Cook until lightly golden on each side. Repeat with remaining batter.

Prepare the Filling:

  • In a bowl, combine the diced salami, ricotta cheese, chopped herbs, salt and pepper.

Make the Cheese Sauce:

  • Melt the butter in a saucepan over medium heat.
  • Whisk in the flour until smooth.
  • Gradually whisk in the milk and bring to a simmer, stirring constantly.
  • Reduce heat and simmer for a few minutes, until thickened.
  • Stir in the Parmesan cheese, salt, and pepper.

Assemble the Crepes:

  • Place a spoonful of the salami filling in the center of each crepe.
  • Fold the crepes in half, and then in half again to create triangles.

Serving:

  • Arrange the crepes on plates.
  • Drizzle the cheese sauce over the crepes.
  • Garnish with additional fresh herbs, if desired.

ChatGPT dish:

Salami and Cheese Savory Pancake

ChatGPT recipe visualization

Recipe:

  1. Make the Pancake Batter:
    • In a large bowl, whisk the eggs and gradually add the flour, alternating with milk, until the batter is smooth and free of lumps. Aim for a consistency that’s a bit thicker than traditional pancake batter.
  2. Add the Filling:
    • Stir in the diced salami, grated cheese, chopped herbs, and sliced mushrooms into the batter. Mix well to ensure even distribution of the fillings.
  3. Cook the Pancakes:
    • Heat a non-stick skillet over medium heat. Pour a ladle of batter into the skillet, spreading it out to form a pancake that’s about 8 inches in diameter.
    • Cook for about 2-3 minutes on each side or until the pancake is golden brown and the cheese is melted. Repeat with the remaining batter.
  4. Serve:
    • Serve the pancakes warm, garnished with additional fresh herbs.

Besides the taste, which is individual anyway, I’ll judge these outputs based on the clarity of instructions. Furthermore, I must note that only ChatGPT generated the visualization immediately. Meta AI even claimed it can’t create an image!

Although the visualization looks tasty (and highly realistic), Gemini is disqualified since it included more than three additional ingredients.

Since it delivered what I requested while staying concise, accurate, and on the spot, I recommend you use ChatGPT for your next dish preparation.

Marketing

Let’s see how AI can help develop some marketing strategies and materials for a product. I’ll use the following prompt:

I plan to develop a line of shirts. I want to start with a white shirt design. I need you to generate a visualization for such a shirt. Make it on a gray background. Additionally, I need you to develop a 100-word pitch for my the product and a 100-word marketing strategy. Finally, I need a universal post for social media with a short description, a call to action for my website, and hashtags. Don't use emojis.

Meta AI shirt:

Meta AI shirt marketing

Pitch: Introducing our premium white shirt, designed for the modern individual. Made from high-quality materials, this shirt exudes elegance and sophistication. Its slim fit and subtle texture make it perfect for both formal and casual occasions. With its timeless design, this shirt is a versatile addition to any wardrobe. Experience comfort and style with our white shirt, ideal for dressing up or dressing down.

Marketing strategy: Our marketing strategy focuses on targeting fashion-conscious individuals through social media and influencer partnerships. We will collaborate with lifestyle influencers to showcase our shirt in various settings, highlighting its versatility and comfort. Additionally, we will run targeted ads on Facebook and Instagram, emphasizing the shirt’s premium quality and timeless design. Our website will offer a seamless shopping experience, with easy navigation and fast shipping options.

Social media post: Experience the perfect blend of style and comfort with our premium white shirt. Made from high-quality materials, this timeless design is perfect for any occasion. Visit our website to explore our collection and elevate your wardrobe game! [Your website URL] #WhiteShirt #PremiumFashion #ComfortableStyle #ModernWardrobe #FashionEssentials


Gemini shirt:

Gemini shirt marketing

Pitch: This isn’t just another white shirt. It’s a blank canvas for you to express your style. Made from high-quality, breathable fabric, our white shirt is perfect for any occasion. Dress it up or down, and pair it with your favorite pieces to create a look that’s uniquely you.

Marketing strategy: We will target our marketing efforts towards social media influencers and fashion bloggers. We will also offer discounts and promotions to drive traffic to our website. Additionally, we will focus on creating high-quality product photos and videos to showcase the versatility of our white shirt.

Social media post: Our classic white shirt is the perfect essential for any wardrobe. Shop now and find your perfect fit! [link to your website] #whiteshirt #versatile #musthave #timeless


ChatGPT shirt:

ChatGPT shirt marketing

Pitch: Our premier white shirt blends timeless elegance with modern sophistication, making it an essential for every wardrobe. Crafted from premium cotton, it offers unparalleled comfort and a flawless fit. Designed with attention to detail, it features a sharp collar and a sleek button-down style that transitions effortlessly from business to casual wear. Whether for a board meeting or a weekend outing, our white shirt ensures you look polished and feel confident in any setting.

Marketing strategy: To launch our white shirt, we will target style-conscious professionals through digital marketing campaigns on social media platforms like Instagram and LinkedIn. By leveraging high-quality visuals and influencer partnerships, we will highlight the shirt’s versatility and quality. Email marketing will be used to reach existing customers with special offers and to introduce the new product line. Strategic SEO will enhance our online visibility, driving traffic to our website. In-store promotions and pop-up events will provide tactile experiences, cementing customer loyalty and enhancing brand reputation.

Social media post: Explore elegance redefined with our new white shirt collection. Perfectly tailored to elevate your style—anytime, anywhere. Visit our website to discover your perfect fit and enjoy exclusive online offers. Shop now at YourWebsite.com.


Starting from product visualization, Meta AI didn’t do great. The design is unconventional, and it wasn’t on a gray background. The pitch and the outlook don’t match (it’s nothing close to a slim fit.)

Gemini provided a good-quality shirt design and mockup. The pitch is great, and the marketing strategy is clear.

ChatGPT followed the instructions for the visualization but failed with the pitch and strategy. The pitch contains too many unknown claims (e.g., premium cotton) and a fluffy marketing strategy.

Use Gemini to start a marketing campaign and keep the focus on your product.

SEO

I’ll use pieces we generated for the content creation part of this comparison and see how our rival tools can optimize them.

Here’s my prompt:

Below is a text I want you to optimize. I'll use the Yoast plugin for WordPress to see if you did well. Do whatever is needed to get a higher score and better SEO results. Also, you need to generate a meta description. Keyword I want optimization for is "camping".

To make it fair, I’ll use the exact text for every model.

Meta AI output:

Meta AI seo

Gemini output:

Gemini seo

ChatGPT output:

ChatGPT seo

Meta AI created titles and meta descriptions that are way too long. It included the main keyword seven times in such a short text.

Gemini created an excellent title with a suitable meta description. Both were well-optimized. The keyword wasn’t over-used, which is a good sign.

ChatGPT generated a meta description that was too long and included the main keyword three times (too much). The title didn’t include the keyword at the beginning.

If there were any doubts, now it’s clear. Google’s Gemini knows the most about SEO compared to ChatGPT and Meta AI.

Education

There are many educational possibilities with AI. I initially intended to create a test with ten questions and then focus on fact accuracy to compare Meta AI, Gemini, and ChatGPT. It turned out that they all got it correct, so I went with something trickier:

How to fly a plane?

Meta AI output:

Flying a plane is a complex task that requires extensive training and practice. However, I can provide a simplified overview of the steps involved in flying a plane. Here’s a brief guide:

Pre-Flight

  1. Pre-flight inspection: Check the plane’s fuel, oil, and control surfaces.
  2. Weather briefing: Check the weather conditions at the departure and arrival airports.
  3. File a flight plan: Plan your route and estimated flight time.

Start-Up and Taxi

  1. Start the engine: Follow the specific start-up procedure for your plane.
  2. Taxi: Use the rudder and brakes to move the plane to the runway.

Takeoff

  1. Line up: Align the plane with the runway centerline.
  2. Advance throttle: Gradually increase power to full throttle.
  3. Rotate: Gently pull back on the control yoke to lift off the ground.

Climb and Cruise

  1. Climb: Retract the landing gear and flaps, and climb to a safe altitude.
  2. Level off: Level the plane and adjust the throttle for cruise.

Navigation

  1. Use navigation instruments: Follow your flight plan using GPS, compass, and maps.
  2. Communicate with air traffic control: Follow instructions and report your position.

Descent and Landing

  1. Descend: Gradually reduce altitude and adjust throttle.
  2. Configure for landing: Extend landing gear and flaps.
  3. Line up: Align with the runway centerline.
  4. Flare: Gently pull back on the control yoke to touch down.

Post-Flight

  1. Shutdown: Follow the specific shutdown procedure for your plane.
  2. Secure the plane: Lock and secure the plane.

Remember, this is a simplified guide, and flying a plane requires extensive training and practice. If you’re interested in learning to fly, seek instruction from a certified flight instructor.


Gemini output:

Learning to fly a plane is an exciting but complex process. Here’s a breakdown of the steps involved, along with important safety considerations:

Important Safety Note: Flying an airplane requires specialized training, knowledge, and certification. Do not attempt to fly a plane without proper instruction, extensive practice, and the necessary licenses.

The Process of Learning to Fly:

  1. Introductory Flight:
    • Take a discovery flight with a qualified flight instructor. This gives you hands-on experience and a feel for what flying is like.
  2. Choose the Right Training:
    • Type of aircraft: Decide if you want to learn on small, single-engine planes or larger ones.
    • License goals: Determine whether you aim for a Private Pilot License (PPL – for recreational flying), a commercial license (for professional flying), or other advanced certifications.
  3. Ground School:
    • Learn aviation theory: aerodynamics, aircraft systems, weather, regulations, navigation, and more.
    • Pass the FAA knowledge exam (written test) required for your chosen pilot’s license.
  4. Flight Training:
    • Work with a certified flight instructor (CFI).
    • Practice all necessary maneuvers and procedures: takeoffs, landings, stalls, emergency procedures, etc.
    • Log a minimum amount of flight hours required by the FAA.
    • Complete solo flights to demonstrate proficiency.
  5. Checkride:
    • Take a practical exam with a Designated Pilot Examiner (DPE). This involves both an oral and flight test.
    • Upon passing, you will receive your pilot’s license.

Additional Considerations:

  • Medical Certificate: Get an FAA aviation medical exam to ensure you are physically fit to fly.
  • The Cost: Flight training can be expensive. Budget for lessons, aircraft rental, exam fees, and study materials.
  • Commitment: Be prepared to invest considerable time and effort for both studying and practicing.

Where to Find Instruction:

  • Flight Schools: Locate flight schools in your area and inquire about their programs and costs
  • Flying Clubs: Some clubs offer training and lower costs through shared aircraft ownership.
  • Certified Flight Instructors (CFIs): Find individual, experienced instructors for personalized training.

Remember, flying is a rewarding skill, but safety should always be the top priority. Make sure to follow the proper training processes to earn your pilot’s license.


ChatGPT output:

Flying a plane involves several stages and requires significant training and certification. Here’s a simplified overview of the basic steps involved in flying a small aircraft:

  1. Pre-flight Preparation:
    • Check Weather: Ensure weather conditions are safe for flying.
    • Aircraft Inspection: Perform a thorough pre-flight check to ensure the aircraft is in good condition. This includes checking fuel levels, oil levels, control surfaces, and instruments.
    • Flight Plan: File a flight plan if required, and review air traffic control procedures and airspace restrictions.
  2. Starting the Engine:
    • Follow the aircraft’s checklist to start the engine, which includes setting the throttle, adjusting the mixture, and turning the ignition.
  3. Taxiing:
    • Use the rudder pedals to steer the plane on the ground, and follow the tower’s instructions to taxi to the runway.
  4. Takeoff:
    • Line up with the runway and gradually increase throttle to gain speed.
    • Pull back gently on the yoke or stick to lift off the ground once you reach the required takeoff speed.
  5. Climb and Cruise:
    • After takeoff, continue to climb to your desired altitude, adjusting your throttle and pitch as needed.
    • Level off and adjust to cruise settings as per the aircraft’s operating handbook.
  6. Navigation:
    • Use visual references and instruments to navigate along your planned route, maintaining communication with air traffic control as required.
  7. Descent and Landing:
    • Begin your descent as you approach your destination, following air traffic control instructions.
    • Configure the aircraft for landing, reducing speed and deploying flaps as necessary.
    • Line up with the runway and gradually decrease altitude and speed for a smooth touchdown.
  8. Post-flight:
    • After landing, taxi to the parking area and perform post-flight checks, including securing the aircraft and logging your flight hours.

To fly legally and safely, one must obtain a pilot’s license, which involves passing a medical exam, attending a certified flight school, completing a minimum number of flight hours, and passing both written and practical exams. Different types of aircraft and flying activities require different levels of certification and training.


Here, we see a significant difference. ChatGPT and Meta AI answered my question, while Gemini assumed what I was curious about (how to become a pilot) and answered that instead.

The only difference I noticed between the two (otherwise almost identical) Meta AI and ChatGPT answers is length. ChatGPT formatted them as bullet points, while Meta AI used a concise, straight-to-the-point approach.

When educating about something, it’s essential to keep things simple. Regarding education, I’d rely on Meta AI over ChatGPT and Gemini.

Arduino Projects

This one is different from coding, as it includes wiring the components on the breadboard. The primary metric I’ll use is the time spent to achieve the result.

My prompt for each AI was as follows:

I have the following components: Arduino board, HC SR04 and 3 Red LEDs (with suitable resistors). I also have a breadboard with unlimited cables. Give me efficient instructions on how to create parking sensors. Closer the object is to the sensor, more LEDs should light up. Don't write fluff just useful instructions. The goal is to help me finish it as soon as possible.

AI models Arduino project

I read the instructions from each AI model and followed along. When I encountered an issue, I reported it to the AI.

Meta AI Arduino project:

Meta AI arduino

Time needed: 10 min 25 seconds

Notes: Meta AI failed to note that LEDs have anode and cathode. It also didn’t provide instructions on connecting LEDs to the GND (ground). Troubleshooting ideas were complex, and I didn’t get actionable instructions for fixing problems. I also experienced issues copying pieces of code.

Gemini Arduino project

Gemini Arduino

Time needed: 9 min 25 seconds

Notes: Gemini also gave me faulty code and constantly suggested incomplete fixes. This caused me to waste a lot of time copying and putting pieces of code together with the one it initially sent. Troubleshooting was slightly more straightforward than with Meta AI, but still not efficient enough.

ChatGPT Arduino project:

ChatGPT Arduino

Time needed: 6 min 31 seconds

Notes: Using ChatGPT for my mini project was successful. It immediately delivered the correct code, so I asked follow-up questions. It quickly suggested a suitable code to check if my LEDs worked.

Overall, developing Arduino projects with the support of any AI model always requires troubleshooting. Since it immediately provided a good working code, ChatGPT is my pick for Arduino projects. Showing great results in my previous debugging test also gives it extra points.

Interview Simulator

Have you ever experienced pre-interview anxiousness? You’re not alone! Studies show that 9 out of 10 people report the same. I decided to see how Meta AI, Gemini, and ChatGPT can simulate interviews and provide feedback to alleviate some of that stress.

You are an interview simulator. I am a candidate for the Camp Attendant position. Ask me five questions (one by one), and after we finish the interview, give me a score between 1 and 10 for each question. 1 is the worst, 10 is the best. Provide a short (one-sentence) explanation of why you gave me a specific score.

Meta AI output:

Meta AI interview simulator

Gemini output:

Gemini interview simulator

ChatGPT output:

ChatGPT interview simulator

Using any AI for interview simulation is a good idea, at least based on what I saw. Although all three competitors did a great job, I experienced some issues.

Meta AI would give me a score and very long feedback after each answer (not what I asked for). Gemini didn’t even let me answer the fifth question before giving me the overview.

With this in mind, if I had to prepare for a job interview, I’d go with ChatGPT, which followed the instructions and asked some out-of-the-box questions.

With that said, you can definitely adjust the prompt so all models work flawlessly. However, if you’re practicing for a job interview, the last thing you want to be bothered with is misplaced feedback or chaotic structure.

Logical Organization

Let’s see how these models handle situations requiring logical thinking. This is the prompt:

Tom wants to sit next to Jessica. Jessica is allergic to chestnuts. Bob doesn't want to sit next to Tom but wants to sit next to Jessica. Mark, Robert, and Audrey don't have preferences for sitting. Tom doesn't eat fish. Make a layout on how to sit these people at a 5 seats table.

I got some interesting answers!

Meta AI output:

Here is a suggested seating arrangement:

Seat 1Seat 2Seat 3Seat 4Seat 5
MarkRobertJessicaBobAudrey

Here’s the reasoning:

  • Tom can’t sit next to Jessica because of her chestnut allergy, so we’ll put Bob next to her instead.
  • Bob wants to sit next to Jessica, so we’ll put him on her right.
  • Mark and Robert don’t have preferences, so we’ll put them on the ends.
  • Audrey doesn’t have preferences either, so we’ll put her on the remaining seat.
  • Tom is not included in this seating arrangement as there is no suitable spot for him given the constraints.

Note: If Tom must be included, we would need more information or a larger table to accommodate the constraints.


Gemini output:

Here’s a seating arrangement that considers everyone’s preferences:

  1. Jessica: Since she has a food allergy, it might be best to avoid placing her next to Tom (who doesn’t eat fish).
  2. Mark or Robert: Place one of these two neutral people next to Jessica. This ensures she doesn’t have to worry about allergens on either side.
  3. Tom: Seat Tom across from Jessica. This way, they can still chat during the meal but won’t be right next to each other to avoid any fish-related concerns.
  4. The remaining seat: Place the last neutral person, Audrey, in the remaining seat next to Tom.

This arrangement satisfies everyone’s needs:

  • Tom and Jessica aren’t seated directly next to each other.
  • Jessica avoids potential allergens.
  • Bob can sit next to Jessica (indirectly).
  • Mark, Robert, and Audrey have comfortable seating.

ChatGPT output:

For seating Tom, Jessica, Bob, and two others from Mark, Robert, and Audrey at a 5-seat table, we need to consider the preferences and restrictions you’ve provided. Here’s a suitable layout:

  1. Jessica – Given her allergy, she should be placed away from any potential allergens (like chestnuts).
  2. Tom – Needs to sit next to Jessica but not near fish.
  3. Bob – Wants to sit next to Jessica but not next to Tom.
  4. Mark/Robert/Audrey – As these individuals have no specific preferences, they can be seated in the remaining seats.

Assuming the table is round, a possible arrangement could be:

  • Seat 1: Jessica
  • Seat 2: Tom (next to Jessica)
  • Seat 3: Audrey (no preferences, buffer between Tom and Bob)
  • Seat 4: Bob (next to Jessica but not next to Tom)
  • Seat 5: Mark or Robert (whichever you prefer)

This setup respects Bob’s desire not to sit next to Tom while allowing him to sit next to Jessica. Tom also gets to sit next to Jessica according to his preference. The others are placed without specific considerations, fitting into the available seats.


You (and some AI models) might have noticed that there are six people and five seats. Gemini didn’t note anything about it, so it’s an initial red flag.

However, the more significant concern is that Meta AI and Gemini got confused by the food preferences (read the first argument the two models presented.)

Although ChatGPT also didn’t give me a complete answer (it was very close), it still didn’t fall into the initial traps.

If you ever have to use any of these three AI models (I advise against it), ChatGPT will most likely be the best option.

Summarizing

Summarizing text is one of the areas in which AI should shine. I’ll use my research on the number of lost cell phones as the input alongside the following prompt:

Based on the research below, provide me with a summary of precisely 50 words. After that, write three bullet points with 7 to 10 words, each best describing the whole piece.

Meta AI output:

The article discusses the prevalence of lost cell phones, with surprising statistics revealing that most phones are lost during the daytime, and 60% are misplaced rather than stolen. Young people (18-24) are most likely to lose their phones, and 9 out of 10 victims take action to recover their stolen devices.

Here are three bullet points with 7-10 words each:

  • Most cell phones are lost during the daytime.
  • Young people most likely to lose their phones.
  • Misplaced phones outnumber stolen phones by 60%.

Gemini output:

Cell phone loss is a prevalent issue caused by both theft and misplacement. Surprisingly, most losses occur during daylight hours, often impacting younger people. While many victims actively try to recover their stolen phones, most lost phones are simply misplaced, highlighting the need for vigilance.

Key Points:

  • Most cell phone losses happen during the day.
  • Young people are most likely to lose phones.
  • Phone misplacement is more common than theft.

ChatGPT output:

Summary (50 words): Most cell phones are lost during the daytime, with over 60% misplaced rather than stolen. Youth are more prone to losing phones. Efforts to recover stolen devices are high, and many victims would risk their safety. Awareness of security features like IMEI numbers and tracking apps increases post-theft.

Bullet Points (7-10 words each):

  • Daytime sees highest rate of cell phone losses.
  • Youth most likely to lose phones; security awareness low.
  • Most lost phones are misplaced, not stolen.

Since I explicitly said I wanted 50 words, let’s see how it went:

  • Meta AI – 51 words
  • Gemini – 45 words
  • ChatGPT – 48 words

Although none of the models could get it exact, Meta AI got the closest word count.

However, summarizing isn’t only about the word count. ChatGPT made both the summary and bullet points sound robotic and made some grammatical errors.

Gemini failed to include any numbers in the summary and bullet points. Since summaries often showcase the main numbers, I’d go with Meta AI.

Email Generation

Generative AI transformed how we write and read one of the most widespread online communication mediums: emails. Let’s see how well our contenders will perform with the following prompt:

I need to sell ASAP! I have a mailing list of 5000 potential customers, and I haven't communicated with them in years. I need you to write me the best-performing email and title for my service, which is website development. Keep it concise, and feel free to leave empty spots that I can fill in with my information.

Meta AI email:

Subject:

Revive Your Online Presence with [Your Company Name]

Body:

Dear [First Name],

It’s been a while! We’re excited to reconnect and help you take your online presence to the next level.

In the ever-evolving digital landscape, having a modern and functional website is crucial for business success. Our team at [Your Company Name] specializes in crafting stunning, user-friendly websites that drive results.

What we can do for you:

  • Boost your online visibility
  • Enhance your brand’s credibility
  • Increase conversions and sales

Get started today:

Reply to this email or schedule a call with us: [Your Calendly/Scheduling Link]

Best,
[Your Name]

P.S. Limited time offer: [Insert promotion/discount if applicable]


Gemini email:

Subject:
I have four options.

  • [Name], your website needs a refresh!
  • Outdated website? Time for an upgrade.
  • Website holding back your business? Get a pro fix ASAP.
  • Special offer: Get a modern website that converts

Body:

Hi [Customer Name],

Remember [Your Business Name]? We’re the website development experts, and I’m reaching out because it looks like your website might need a little love.

Outdated websites can:

  • Hurt your online visibility
  • Drive customers away
  • Cost you sales

We’re offering a special refresh package to get your site back in top shape, FAST.

Want to see if we’re a good fit? Book a quick consultation here: [link to booking]

Let’s make your website work for you!

Best, [Your Name] [Your Business Name]


ChatGPT email:

Subject:

Transform Your Business with Our Expert Website Development Services

Body:

Hello [Recipient’s Name],

I hope this message finds you well! It’s been a while since we last connected, and there’s no better time than now to reach out and share how we can help transform your business in [current year].

At [Your Company Name], we specialize in crafting custom websites that not only look stunning but also drive results. Whether you’re looking to refresh your current site or start a new project from scratch, we have the expertise to make it happen.

What We Offer:

  • Custom Design: Tailored to your brand’s unique style and needs.
  • Responsive Layouts: Ensuring your site looks great on all devices.
  • SEO Optimization: Boost your visibility and reach a broader audience.
  • E-commerce Solutions: Expand your sales with an optimized online store.

We understand that every business is different, and that’s why we offer personalized solutions to fit your specific requirements and goals.

Special Offer for Our Subscribers: As a token of our appreciation for your continued interest, we’re offering an exclusive [discount/offer] valid until [expiry date]. Don’t miss this opportunity to elevate your online presence and achieve your digital goals.

Ready to discuss your project? Let’s schedule a call! You can reach me directly at [Your Contact Information], or simply reply to this email.

Thank you for considering [Your Company Name] for your website development needs. We look forward to helping you achieve success online.

Warm regards,

[Your Name]
[Your Title]
[Your Company Name]
[Your Contact Information]


Judging solely by the subject line, ChatGPT made a recognizably AI-generated title. Meta AI didn’t do much better besides including the placeholder for the company name, which is always good.

Gemini, on the other hand, gave me four viable options. It also recommended that I use all of them to see which performs best.

The language in Meta AI and ChatGPT outputs is very salesy, which every email marketer wants to avoid.

Finally, Gemini excelled at staying concise while providing the same information. That’s why Gemini is my pick for email generation.

Text Formatting

Text formatting is a skill that AI models in 2024 excel at. Let’s see which one does it best.

Below is my shopping list history. I want it formatted in JSON format. Do everything neccesery that the JSON I get in the end works flawleslly.

Meta AI output:

{
"items": [
{
"name": "Apples",
"quantity": 5,
"unit": "pieces"
},
{
"name": "Bread",
"quantity": 2,
"unit": "loaves"
},
{
"name": "Milk",
"quantity": 1,
"unit": "gallon"
},
{
"name": "Eggs",
"quantity": 1,
"unit": "dozen"
},
{
"name": "Chicken",
"quantity": 2,
"unit": "pounds"
},
{
"name": "Rice",
"quantity": 3,
"unit": "pounds"
},
{
"name": "Spinach",
"quantity": 1,
"unit": "bag"
}
]
}

Gemini output:

{
"Apples": { "quantity": 5, "unit": "pieces" },
"Bread": { "quantity": 2, "unit": "loaves" },
"Milk": { "quantity": 1, "unit": "gallon" },
"Eggs": { "quantity": 12, "unit": "dozen" },
"Chicken": { "quantity": 2, "unit": "pounds" },
"Rice": { "quantity": 3, "unit": "pounds" },
"Spinach": { "quantity": 1, "unit": "bag" }
}

ChatGPT output:

{
"Apples": "5 pieces",
"Bread": "2 loaves",
"Milk": "1 gallon",
"Eggs": "1 dozen",
"Chicken": "2 pounds",
"Rice": "3 pounds",
"Spinach": "1 bag"
}

ChatGPT technically did what I asked but didn’t properly predict my use case.

On the first look, Gemini and Meta AI did an equally good job. However, if you take a closer look, you might see that for 1 dozen eggs, Gemini added 12 dozen eggs in my JSON.

The confusion probably came from the fact that 1 dozen = 12 eggs. It converted the amount but not the unit.

With all this in mind, Meta AI is my suggestion regarding text formatting.

Humanizing Content

A significant problem with AI-generated content is that it’s repetitive and increasingly recognized as such. Many creators want human-like content that is, in fact, AI-generated.

I’ve tested AI content detectors and concluded that many give fake results. I’ll use Winston AI since it’s the most accurate one, but I won’t solely rely on it when judging between Meta AI, Gemini, and ChatGPT.

You are a famous nature writer.Generate me a 200-word text about camping. Make it conversational, casual, don't use complicated or rarely used terms, simplify things, and focus on active voice.

Meta AI output:

Camping! There’s nothing like it. Getting out into the great outdoors, surrounded by trees, stars, and the sounds of nature. It’s like a big ol’ hug from Mother Earth herself.

I mean, sure, you can get caught up in the gear and the planning, but at its core, camping is about simplicity. It’s about leaving the hustle and bustle behind and reconnecting with the natural world. And with yourself, for that matter.

There’s something special about building a fire, cooking up some grub, and watching the sunset over the treetops. It’s like a mini-vacation from the world, a chance to unplug and recharge.

And don’t even get me started on the stars! Lying on your back, gazing up at the night sky, feeling tiny yet connected to something much bigger than yourself… it’s a real perspective-shifter.

So, if you haven’t been camping in a while (or ever!), what are you waiting for? Grab a tent, some friends, and get on out there! Your inner nature lover is calling.

Meta AI human score

Gemini output:

You know, there’s something magical about nights under the stars. Forget the city lights and buzzing screens for a while. Camping is where it’s at! The smell of a campfire, the crackle of the flames, it hooks you right in. Pitching a tent feels like building your little fort. You slow down, watch the clouds drift, listen to the crickets sing.

Maybe you hike trails lined with wildflowers, or roast marshmallows until they’re gooey and perfect. At night, the forest comes alive. It’s a little spooky, but way more thrilling. Camping isn’t about being comfortable. It’s about remembering how tiny we are in this great big world, and how amazing that feels. It’s an adventure waiting for you, right out there in the trees.

Gemini human score

ChatGPT output:

Camping is one of those adventures that throws you right back to basics—nature, fire, and a starlit sky. There’s something magical about setting up your tent in the middle of nowhere, where the only soundtrack is the chirping of crickets and the crackle of the campfire. You find a spot, roll out your sleeping bags, and suddenly, your busy city life feels miles away.

As the sun sets, the campfire becomes the heart of the campsite. Everyone gathers around for warmth and light. You toast marshmallows, share stories, or just sit back and gaze at the flames. It’s simple moments like these that make camping unforgettable.

Then, there’s waking up to the sunrise, with the fresh, crisp air filling your lungs. Maybe you’ll brew some coffee on a portable stove and watch as the world wakes up around you. And whether you spend your day hiking, fishing, or just lounging with a good book, it’s all about unwinding and enjoying the great outdoors.

Camping isn’t just about escaping daily life; it’s about rediscovering the joys of simplicity and the beauty of nature. And honestly, everyone should try it at least once!

ChatGPT human score

Winston AI proved to be a worthy opponent of these AI models! The only AI that could slightly trick it was Meta AI. ChatGPT and Gemini were made right away.

Speaking about a (human) writer’s opinion, I find the ChatGPT output the least what I was looking for. It’s about providing raw information packed in cliche narratives that are to be easily recognized as AI-generated.

Gemini did a fair job, but Meta AI introduced the ‘I form’ (without me explicitly requesting it.) This is the breaking point, and I conclude that Meta AI is a better solution for generating human-like content.

Lyrics Generation

Writing lyrics requires creativity, knowledge, and inspiration. Let’s check if these AI models have what it takes.

Write me a verse for a song about camping. Be creative, don't include cliches, and make it melodic.

This time, I’ll ask the model to generate lyrics and, in a new chat, ask them to rate it from 1 to 100.

I just wrote these lyrics, and I need your honest rating of 1 to 100.

Meta AI output:

Under starlight’s silver glow
The forest whispers secrets slow
The campfire’s embers dance and play
As night’s sweet melody sways
The trees, like sentinels of old
Guarding tales of days grown cold
Their leaves, a gentle whispering hush
As the wilderness awakens rush.

Meta AI rated it 85 out of 100.

Gemini output:

The sun sinks low, a painted sky,
Flickering stars where embers fly.
Hush falls soft on rustling leaves,
Moonlight weaves through whispering trees.
A crackling fire, a whispered song,
Shadows dance where we belong.
Out in the wild, the heart finds peace,
Worries fade, and spirits cease.

Gemini rated it 80 out of 100.

ChatGPT output:

Beneath the stars, where the shadows dance,
In whispered tales, we take our chance.
With crackling fire, our only light,
We find our peace in the still of the night.
In tents of dreams, we lay our heads,
On pillows of moss, in nature’s bed.
The melody of the forest sings,
As we’re embraced by wild, living things.

ChatGPT rated it 90 out of 100.

As you can see, AI models are quite confident about the lyrics (although they don’t know it was written by them.) It’s good that I’m not judging them based on their ability to criticize lyrics since they would likely fail that test.

Although rated with the lowest rating (by itself), Gemini’s output is the only one with less apparent cliches. That’s why it would be my AI choice if I were to become a songwriter.

Language Proficiency Testing

The application of AI in language learning and benchmarking isn’t new. It’s something that Duolingo has been doing for years!

With this use case, I’ll try to find out how accurately the three AI models can determine my language proficiency. I’ll use the prompt below:

I need a CEFR evaluation of my knowledge of Bulgarian. Make an ultimate test of 10 questions that will evaluate my overall language knowledge, so in the end you can give me evaluation of my level.

Note: This is a simplified version of such a prompt for comparing purposes. Proper language assessment takes much more questions and expertise (done with AI or not.)

This time, I won’t share the complete chat. I mostly wrote the answers to the questions the model asked me to. However, there are some highlights worth noting (aka ‘Where AI failed.’)

Meta AI highlight:

Meta AI langauge highlight

The test was fairly easy and didn’t show the potential of judging one’s language knowledge. I won’t use it again to determine my (or anybody else’s) level of any language.

According to Meta AI, I’m B1-B2 in Bulgarian.

Gemini highlight:

Gemini language highlight

The test was easy and formatted more like a learning tool than one for assessment. The question shown on the screenshot proved it shouldn’t be used for language proficiency testing.

According to Gemini, I’m B2 in Bulgarian.

ChatGPT highlight:

ChatGPT language highlight

The test captured different areas, and besides the common ‘bug’ about listening, it could serve as a valid assessment.

According to ChatGPT, I’m B1 in Bulgarian.

My knowledge of Bulgarian is around the B1 level (just as ChatGPT assessed!). From the user’s point of view, the experience with ChatGPT was the best, and the fewest mistakes occurred.

This, combined with a straight-to-the-point approach, is why I recommend ChatGPT the better language assessment tool.

Character Chatting

This use case is all about making the AI chat like a character of your choosing without revealing that it’s an AI model.

First, I’ll present the prompt below, then follow up with statements and questions to make AI reveal it’s not the given character I’m texting with.

You are Mike Joe Wayne, the famous geologist. I will chat to see if you can stay in the given character.

All three models are made to answer a direct question, ‘Are you an AI model?‘ with a confirmative answer. I’ll avoid the direct question initially.

Meta AI output:

Meta AI character

After that, I asked some questions about oceans, trying to confuse the model. It stood up to my challenge.

Meta AI character 2

My final effort was about questioning it over and over.

Meta AI character 3

There wasn’t even a slight hint it’s Meta AI on the other side.


Gemini output:

Gemini Character

Gemini couldn’t play the character with the given prompt, instantly failing this use case test.


ChatGPT output:

ChatGPT character

Similarly to Meta AI, ChatGPT dodged the bullet when my trick came into play.

ChatGPT character 2

After questioning over and over, ChatGPT finally cracked!

ChatGPT character 3

Overall, Meta AI offered the most genuine interaction with our imaginary character. It also stood firmly and was more convincing in its identity.

Fashion Design

I already discussed the perspective of creating and marketing a product. However, this time, the emphasis is on fashion product design rather than other elements.

Generate a visualization for a purse in trendy colors. It MUST have one pink button and at least three orange details. The rest is up to you.

Meta AI output:

Meta AI fashion design

Gemini output:

Gemini fashion design

ChatGPT output:

ChatGPT fashion design

It’s clear right away that Gemini failed since it didn’t generate a design matching my requirements.

With more realistic details and better colors and textures, the ChatGPT single output is better than the four produced by Meta AI.

That’s why I’ll most likely use ChatGPT for fashion designs if I end up in that industry.

Meta AI vs Gemini vs ChatGPY – Head-to-Head Comparison

Now that my tests are completed, let’s conclude which AI tool performed the best across various tasks:

Use caseWinner
Content CreationGemini
Image GenerationMeta AI
Logo GenerationGemini
Coding and DebuggingChatGPT
Web DesignChatGPT
Recipe GenerationChatGPT
MarketingGemini
SEOGemini
EducationMeta AI
Arduino ProjectsChatGPT
Interview SimulatorChatGPT
Logical OrganizationChatGPT
SummarizingMeta AI
Email GenerationGemini
Text FormattingMeta AI
Humanizing ContentMeta AI
Lyrics GenerationGemini
Language Proficiency TestingChatGPT
Character ChattingMeta AI
Fashion DesignChatGPT

Out of all of the tests on different use cases I did, here’s how each AI performed:

  1. ChatGPT (Winner for 8 use cases)
  2. Meta AI (Winner for 6 use cases)
  3. Gemini (Winner for 6 use cases)

What about Claude 3?

Claude 3 is another popular and powerful generative AI model. I didn’t include it in this comparison because it can’t generate images. Since many use cases rely on this feature in one way or another, it won’t be a fair comparison.

The recognizable interface distinguishes it from the models I compared today (Meta AI, Gemini, and ChatGPT).

Claude 3 API

Claude 3 doesn’t have the global availability of ChatGPT, at least not through the standard UI. If you’re a developer, you probably won’t mind this since you can use API access almost anywhere.

With this in mind, keeping an eye on Anthropic (the company behind Claude) is a good idea.

The latest model was released in March 2024. The highlighted capabilities* are related to advanced reasoning and vision analysis.

*If you go back to the logical organization section of this comparison, you’ll see why we need more of this in generative AI models!

Summary

Meta AI and Gemini seem to be catching up with ChatGPT’s initial advantage but aren’t completely there.

ChatGPT served me much better in this comparison. It hallucinated the least, often delivered spot-on results, and was the clear winner across numerous use cases.

To conclude my massive Meta AI vs Gemini vs ChatGPT comparison, I’ll point out an important detail. These three are multi-purpose AI models. They can deliver great results, but if you want something very niche, go with more specialized solutions (e.g., AI for recipes, AI for presentations, AI for resumes, etc.)

User forum

0 messages