Le coût réel des générateurs d'apps IA (j'ai créé la même app dans tous)

Are AI app builders actually saving you money or are they just expensive marketing hype? Because look, everyone's talking about how AI can build apps for pennies on the dollar. But when I actually try to build a real functional app on some of these so-called budget-friendly platforms, [music] ended up costing me hundreds of dollars. So I decided to put this to the test. I took four of the most popular AI app builders, the ones that everyone's raving about, and built the exact same app using the exact same prompts on each platform. But what I discovered about the real cost will probably shock you because the cheapest option definitely wasn't what you'd expect. I'm going to break down every single cost, hidden fees, the upgrade traps, the actual functionality you get for your money, so that by the end of this video there will be a clear winner. Plus I'll show you the exact prompts that I used so you can replicate this test yourself, too. So stick around until the very end to see which AI app builder takes the crown. Here's the exclusive part. I've also created a free course showing you how to build apps and websites and even SaaS products with [music] the winning tool, all completely without any code. Now this course normally costs $299 to join, but for people watching this video, thank you very much, it's completely free. Now this isn't just theory, you'll learn how to create real profitable applications using AI. You can only access this course by watching until the end, so please don't skip ahead. Check out the link in the description if you just can't wait. I got you. Building apps with AI sounds cheap first until you actually start using it. Every time something breaks, every time you have to re-prompt, you're spending more money, and that adds up fast, especially when you're trying to build something with real features instead of just a demo. >> [music] >> So I treated this more like a report card for each platform. And the goal here is simple, we're building the same card collecting app across all platforms, and then tracking how much it actually costs to get there. [music] And each one has to handle the same core features as well. Login and authentication, a database, and a Stripe-based [music] premium system, so everything stays fair, and everything is judged on three things. First is whether the prompt actually works. [music] Every feature starts with the exact same prompt, but if the output is broken or incomplete, well, I have to run it again, and that's [music] important because in real use you don't just pay once, you pay for every single attempt it takes [music] to fix it. Second is cost, which is the main focus of this test. Every prompt uses up credits, [music] tokens, or actual money depending on the platform. So if something takes multiple tries, the cost increases quickly. The platforms that get things right in fewer [music] prompts have a clear advantage here. And lastly, overall value. This is where everything comes together, >> [music] >> and it's not just about which one is the cheapest per prompt, but which one actually gives you working features without wasting credits. So because a platform might look affordable at first, but if you keep re-prompting just to get something usable, well, it just ends up being more expensive [music] than expected. So instead of just looking at which app looks better, this really does come down to efficiency. How many attempts it takes, how much it costs, and whether it actually works in the end. For the first app, we're starting with Bolt. Now this is the first platform in the test, so it basically sets the baseline for everything that comes after. Same prompts, same app, same expectations, so we can clearly see how each platform performs under the exact same conditions. Now Bolt starts with 15.1 million tokens, and its pricing works out to $25 for every 10 million tokens. And that gives us a clear way to track how much each feature actually costs as we go. For the foundation, this is the exact prompt we're going to use. Create a card collecting web application with a modern design. The cards we'll be collecting are varied. It can range from basketball, baseball, MTG, Pokémon, etc. Include a homepage with navigation, user dashboard, and side menu. Do not build any functionality yet, only create your idea of the app's design. Make it responsive and mobile-friendly. And Bolt handles this in a single prompt with no revisions, and it used around 200,000 tokens, bringing the total down to 14.9 million remaining. Now in terms of cost, that comes out to about 50 cents for this step. [music] The result is pretty solid. Bolt generated a clean, professional-looking layout with proper placeholders in around 3 minutes or so. For a foundation prompt, this is exactly what you want here. It's quick, it's straightforward, and it's very cost-efficient. [music] Next is the actual card collecting feature, where things really do start to matter more since we are moving from design into real functionality. Now here's the prompt we're going to use. [music] Remove all placeholder data on the platform, then build the actual card adding functionality. Allow users to add cards to their collections, set details, filters, and all necessary info for their collection. Afterwards, build the card collection gallery. Set this up in a proper database. Whenever a user adds a card, their card should be listed here and should be viewable. And Bolt handles this in one prompt again with no revisions needed. That's great. [music] And it used around 500,000 tokens, bringing the total down to 14.4 million remaining. Now that puts the cost of this feature at about a dollar and a quarter, and honestly, this is a good result. Bolt was able to build both the card adding system and the gallery properly, and everything worked right right away. Next is the viewing modal, which is a smaller feature, but still important for usability. Now for this step, we prompted Bolt to add a modal that opens when a user clicks on a card showing the full image, detailed stats, a description, and of course a delete option. Bolt handled this in a single prompt with no revisions. It used around 300,000 tokens, bringing the total down to 14.1 million remaining. And that puts the cost of this feature at about 75 cents. No issues here. The modal works as expected, displays the right information, and responds properly when interacting with cards. It's another straightforward win here. One prompt, low cost, and no problems. Adding authentication is where things start to get a bit more serious because up to this point, everything has been pretty straightforward, but now we're dealing with real users, real data, and making sure that everything is actually connected the way it should be. The prompt here is to add login and sign up, connect everything to a database, and make sure that each user only sees their own collection. Now Bolt is able to get it working, but not on the first try. I have to run the prompt twice before everything works properly. In total, it uses around 800,000 tokens, bringing the remaining balance down to 13.3 million. And that puts the cost for this feature at about $2. In the end, everything does function correctly. Authentication works, the database is connected, and our users only see their own data. But this is the first time a retry is needed, and you can already see the downside here. One extra prompt instantly doubles the cost for the same feature. And this is where things start to just break down. Here we'll ask Bolt to integrate Stripe, set up a premium subscription system, and add an AI scanning feature locked behind that subscription. And Bolt struggles a lot with this. It takes over 10 prompts with multiple revisions and trial and error, and even after all of that, it still doesn't fully work. At this point, it has to be considered a failed feature. In total, it consumes around 3.6 million tokens, bringing the remaining balance down to 9.7 million. And that alone costs about $9 just for this one feature. And that is a huge problem. Not only does Stripe not work properly, but Bolt also doesn't have built-in AI capabilities. So the scanning feature isn't properly implemented, either. And to even make that part work, you'd need to bring your own API key, which adds more complexity outside of the platform. Now technically, yeah, you could keep prompting and try to fix everything, but at this point we've already spent a lot just trying to get it working. So for this test, I'm just going to stop here and move on. This is easily the weakest part so far. High cost, lots of retries, and no working result. After that, we're going to move on to the collection and wish list sections. We'll prompt Bolt to build out the my collection and wish list features along with their full functionality. And Bolt handles this in a single prompt with no revisions. It uses around 400,000 tokens, bringing the remaining balance down to 9.3 million. And that puts the cost of this feature at about a dollar. No issues here. Both sections are implemented properly, and everything works as expected. It's a pretty straightforward feature, and Bolt delivers it cleanly. One prompt, no fixes, and at a reasonable cost. Next up is the market watch section, which is a bit more data-heavy compared to the previous features. [music] What we're going to ask here is for Bolt to create a section with a large directory of active rare cards that users can explore, and to integrate wish list functionality into it. And Bolt completes this in a single prompt. It uses around 600,000 tokens, bringing the remaining balance down to 8.7 million. And that puts the cost at about $1.50. The feature works well, the directory is there, the structure makes sense, >> [music] >> and it does what it's supposed to do without needing any more fixes. And so it's another clean one prompt implementation. So the final major feature is the marketplace, and for this one, everything has to be done in a single prompt with no testing or revisions. Now we're going to ask Bolt to build a full marketplace system, separate login and sign up, users creating listing, and other users being able to view them, contact sellers, apply filters, and even include demo listings for testing. >> [music] >> And Bolt manages to complete all of this in just one prompt, and it used up around 900,000 tokens, bringing the remaining balance down to 7.8 million. And that puts the cost at about $2.25. And surprisingly, it got everything in place. All the required pieces are there, from listings to filters to user interaction, without needing any follow-up prompts. Now for a feature this large, getting it done in one shot is a solid result. Overall, Bolt had mixed results throughout the build. For simpler features like the foundation, card gallery, and viewing model and performed really well. Everything was done in single prompts, sustained within a low cost range of about $0.50 to $1.25, and worked without any issues. But, once things got more [clears throat] complex, well, that's where problems really started showing up. The biggest example is the Stripe and AI scanning feature, which took over 10 prompts, cost around $9, and still didn't end up working properly. Authentication also needed two attempts, which already doubled what should have been a simple one-prompt feature. >> [music] >> So, while Bolt handled basic UI and straightforward functionality pretty efficiently, it struggled with more advanced integrations. And in total, it used around 7.3 million tokens, which comes out at about $18.25, and that's with one major feature still broken. On top of that, the lack of built-in AI support makes it harder to implement more modern features without relying on external APIs, which then adds just even more friction. Now, in terms of pricing, Bolt keeps things pretty straightforward with a token-based system. There's a free plan with a 300,000 daily limit and 1 million tokens per month, which is really only enough for testing or very small projects. The Pro plan is $25 a month for 10 million tokens, and then the Teams plan is $30 per user with the same token allocation, but added collaboration features. Enterprise is custom depending on what you need, and at first glance, the Pro plan actually looks pretty affordable. But, the issue is how quickly those tokens get used, especially when features just don't work on the first try. And the pricing only feels cheap if everything works in one prompt, which, as we saw, is not always the case. For this build, Bolt used a total of 7.3 million tokens across more than 18 prompts, and that brings the total development cost to about $18.25. And breaking it down, most of the simple features stayed within a reasonable range around $0.50 to $2.25 each, and worked without issues. But, the Stripe and AI scanning feature alone cost $9 and still doesn't work, which ended up being nearly half the total cost, and that's really the main problem here. One complex feature ended up eating a huge portion of our budget without even delivering a working result. So, even though the Pro plan gives you 10 million tokens for 25 bucks, the single project already used about 73% of that, and that's with a major feature still broken. And at that point, you'd either need to buy more tokens or upgrade your plan just to keep going. So, overall, Bolt can be cost-efficient for simple features, but once you start dealing with more advanced functionality, well, then the cost becomes unpredictable very quickly. And when retries really start stacking up, it just stops being a budget-friendly option. So, for our second app today, we're moving on to Replit, and just like before, we're using the exact same prompts from Bolt so that everything stays consistent and comparable across the board. Now, Replit works a bit differently when it comes to pricing. Instead of tokens or credits, it directly shows you the total cost of the app as you build, and that actually makes things a bit easier to track since we can see exactly how much each step adds to the total. So, instead of counting tokens here, we measure the cost per feature by looking at the difference between each step. And, for example, if the total cost goes from $1 to $1.25, then that feature cost $0.25. It's a much more straightforward system, and it gives us a clearer picture of how much we're actually spending as we go. So, for the foundation here, Replit handles it in a single prompt with no revisions at all. The total app cost after this step is about $0.74, which means the foundation itself cost $0.74. The result is clean. Replit generates a professional-looking card collecting app with a solid structure and proper placeholders right away. >> [music] >> The layout is organized, and everything is placed where, well, we'd expect them to be. Now, one thing that stood out here is how straightforward the pricing is. There's [music] no need to convert tokens or credits. What you see, well, is exactly what you get and what you pay for. So, for a single prompt, $0.24 for a working foundation is a pretty solid start. Moving into the card collecting feature, this is where the app starts handling real data. Replit completes this in a single prompt with no revisions, and now the total app cost moves up to $1.86, which means this feature cost about $1.12. And the result works well. The app now includes card adding functionality along with browsing and filtering in a proper gallery view. Everything is structured correctly, and the data setup makes sense for a real collection system. Again, the pricing stays clear. Replit directly charges $1.12 for this feature, and since it works in one prompt, there's no extra cost from retries. For the viewing model, this adds another layer of interaction to our app, and Replit handles this in one prompt again with no revisions. And so, the total app cost increases to $3.77, which means this feature cost $1.91. And the result works as expected. The model opens up properly when a card is clicked, and it shows us all the necessary details, the image, the stats, and the collection info. And everything is still responsive and behaves the way it should. And the cost here is noticeably higher compared to the previous features, but since it does still work in one prompt without any fixes at all, it does stay predictable as well. Next, we add login, authentication, and database integration in this step, and this is where the app starts handling actual users and personal data. Now, Replit needed two prompts here with one revision before everything actually finally worked properly. [music] And the total app cost moves up to $4.78, which means this feature cost $1.01. In the end, the result works. Login and sign up are in place, the database is connected, and users only see their own data as expected. Even with the extra prompt, the cost stays relatively low compared to the earlier features. So, while it didn't get it right on the first try, it still ends up being a fairly efficient implementation overall. So, now let's go ahead and add Stripe, the premium subscription system, and the AI scanning feature. This is easily one of the most complex parts of our build. And Replit needed six prompts here with multiple revisions before everything worked properly. The total app cost jumps to $19.93, which means this single feature cost $15.15. That's a huge increase compared to everything before. And the good part though is that it does work in the end. Stripe is properly integrated, the checkout flow functions properly, and premium features unlock as expected. >> [music] >> The AI scanning feature is also in place, but the cost to get there is high. Needing six prompts for one feature significantly increases the total, and this ends up being the most expensive build by far. So, here we're adding more structure to the app by introducing [music] the collection and wish list sections so that our users can actually organize their cards instead of just viewing them in one place. And it works right away on the first try, and the total app cost jumps to $24.49, so this feature ends up costing $4.56. And everything functions the way that we'd expect. We can organize our cards into collections, manage a wish list, and move things around without any problems. The only thing that really stands out here is the price. Even though it works immediately, $4.56 feels a bit high for something this straightforward. After setting up collections and wish lists, we can now move into the market watch section, which shifts the app from personal tracking to something more data-driven. And this goes through in one prompt with no revisions, and the total app cost increases to $26.62, so this feature adds $2.13. [music] And this section is fully usable right away. Users can browse a wider range of cards, explore market data, and use that information alongside their own wish list. >> [music] >> And the layout also feels more structured enough to handle a large directory without it really becoming messy. Compared to the previous step, this feels more reasonable in terms of cost. [music] It's still not cheap, but at least it works right away and doesn't require any extra attempts. For the final feature, this is where everything comes together with the marketplace system. And this includes creating listings, letting users view and filter them, contact sellers, and adding a separate login flow along with demo data. All done in a single prompt with no testing or revisions. And Replit handles this in one go. The total app cost reaches cost about $4.61. Everything is implemented as expected. The marketplace is functional, listings are visible, and user interactions are in place without needing any follow-up prompts. Now, for something this large though, getting it done in one shot is a pretty strong result. The cost is a bit on the higher side, but at least it doesn't require multiple attempts to get working. And just like Bolt, Replit shows us mixed results throughout the build. On the simpler side, it handles things pretty well. Features like the foundation, the card system, and the model worked in single prompts, and the cost stayed relatively controlled, ranging from about $0.74 to $4.61 depending on the feature. But, once we got into more complex parts, well, that's where the cost started to spike. The biggest example is the Stripe and premium system, which took six prompts, [music] and ended up costing $15.15 all on its own, almost half of the total budget. Authentication also needed a second attempt though, but it did stay relatively cheap at a dollar and a penny. And one thing Replit does really well is pricing clarity. We always see exactly how much each step cost, which makes it easier to track spending without dealing with tokens or any other conversions. The downside is that when features don't work on the first try, well, then the cost climbs quickly. [music] Basic functionality is reliable, but more advanced integrations do take multiple attempts, and that's where things start getting expensive. So, looking at pricing here, Replit keeps things very straightforward with a pay-as-you-go model. There's a free starter plan with limited daily credits, mainly for testing. The Core plan is $20 a month and gives you $20 in credits, while the Pro plan scales that up to $100 a month with more access to advanced features [music] and models. Then there's Enterprise for larger teams with custom setups, and the biggest advantage here is clarity. Every prompt has a direct dollar cost, so you always know exactly how much you're spending without ever needing to convert tokens or credits. But that [music] same system is also where things can get tricky. Costs are easy to track, but they're not easy to control. If a feature takes multiple attempts, the price climbs up quickly, and we saw that clearly with the Stripe integration alone costing $15.15. So, for this build, Replit used 14 prompts in total, bringing the full development cost to $31.23. And breaking it down, simpler features stayed fairly reasonable, most of them falling between $0.74 and around $4.61. But once again, the Stripe and premium system stands out as the biggest cost driver, taking up nearly half of the entire budget. And that's really the main issue here with Replit. Even though pricing is transparent, [music] it's still unpredictable when things just don't work on the first try. And with a core plan only giving us $20 of monthly credits, this single app already goes over that limit by $11.23. That means you either have to buy more credits or just move up to a higher plan just to finish this project. Overall, Replit makes it very easy to understand what you're paying for, but when it comes to complex features, then the cost can escalate quickly, and that makes it less practical if you're trying to stay within a fixed budget. For the third app, we're now testing Lovable, and again, we're using the same prompts from the previous builds. Lovable starts with 91.8 credits, and pricing is based on credits instead of tokens or direct cost per prompt. The pro plan gives us 100 credits for $25 along with additional daily credits, so we can track how efficiently each feature uses those credits. So, for the foundation, Lovable handles it in a single prompt with no revisions. It uses about 2.3 credits, bringing the total down to 89.5 remaining, and that puts the cost for this step at around $0.58. And the results look good. Lovable generated a clean, dark-themed card collecting app with a polished layout and proper structure in place. >> [music] >> It takes around 4 minutes to complete, and for a single prompt, the output feels well put together still and efficient in terms of credit usage. Moving into the card collecting feature, this is where Lovable starts to stand out a bit. It handles [music] this in a single prompt with no revisions, using about 4.6 credits, and that brings the total down to 84.9 remaining, and the cost for this feature comes out to around $1.15. And the feature works as expected. Users can add cards, they can view them in the gallery, and of course, manage their collection properly. But what's really interesting here is that Lovable goes a step further. Without being asked, it also sets up login and authentication during the step, and that's a big advantage here because it saves us both prompts and credits that would have been needed later on. So, that kind of behavior shows a better understanding of what the app actually needs [music] instead of just doing exactly what the prompt says. Adding the viewing modal is straightforward here as well. Lovable handles this in one prompt with no revisions, using about 2.6 credits, and that brings the remaining balance to 82.3, and the cost for this feature coming out to around $0.65. Everything works cleanly. The modal opens up when a card is clicked and shows the full details, the stats, and collection info without any issues. It's another smooth one, low cost, no retries, and no complications. Now, for the login, authentication, and database setup, there's actually no actual work needed here. Lovable already handles that earlier during the card collecting feature, even though it wasn't even explicitly asked for at that stage, and that's a huge plus because again, it saves us an entire feature prompt and keeps the total cost lower. So, now let's go ahead and move into Stripe, the premium access, and the AI scanning feature, which is usually one of the more complex parts. And that's exactly why Lovable needed three prompts here with a couple of revisions before everything worked properly. In total, it used about 6.8 credits, bringing the remaining balance to 75.5, and that puts the cost for this feature at around $1.70. After the revisions, everything functions correctly. The checkout flow works, users can subscribe, and premium features unlock as expected. The AI scanning feature is also integrated properly. Compared to what we saw earlier with other platforms, this is a strong result. It still needed a few attempts, but the total cost stays relatively low, and more importantly, the feature actually works in the end. All right, so let's go ahead and add the collection and wish list sections, giving users more control over how they organize their cards. Lovable completes this in a single prompt with no revisions, using about 4.1 credits, and that brings the remaining balance to 71.4, and the cost for this feature coming out to around $1.03. Everything works smoothly. Our users can separate cards into collections and manage their wish list without any issues. And the structure also feels clean and easy to use. Market watch section adds a different layer to the app by bringing in a broader card data and trends. [music] And Lovable handles this in one prompt, using about four credits. That brings the remaining balance to 67.4, and the cost for this feature coming out to about $1. The section is fully functional. Users can explore card values, [music] track trends, and use that information alongside their own collection and wish list. And for the final feature, the full marketplace system, which ties everything together into a complete app experience. And Lovable completes this in a single prompt, using about 6 and 1/2 credits. That brings the remaining balance down to 60.9, and the cost for this feature coming out to around $1.63. Everything is built in one go. The listings, the user interactions, the filters, and the overall marketplace structure are all in place without needing any revisions. For a feature this large, though, getting it done in one prompt like this is a pretty strong result. Lovable is easily the most consistent platform so far. Across the entire build, it delivers every feature successfully with very few issues. Most features are completed in a single prompt, and even the more complex ones like Stripe and premium access, those only take a few attempts before working properly. One of the biggest advantages here is how it handles things proactively. It already sets up login and authentication earlier in the process without being asked, [music] which saves us both time and credits. Now, that kind of awareness makes the whole build feel smoother. In total, Lovable uses about 30.9 credits, which comes out to around $7.73 across just nine prompts. And more importantly, everything actually works by the end. The combination of low cost, fewer retries, and consistent results makes this a very reliable option for building apps like this. So, now let's talk about pricing. Lovable uses a credit-based system that's pretty easy to follow. There's also a free plan with five daily credits, which is enough for small tests or basic projects. The pro plan is $25 a month for 100 credits plus daily bonuses that can't bring it up to 150 credits if used consistently. >> [music] >> The business plan is $50 with the same credit allocation, but adds more team-focused features, and enterprise is flexible depending on the setup. >> [music] >> So, one thing that stands out here is that credits are shared across users, which is great for teams, but can also mean usage can vary a lot if multiple people are building at the same time. So, for this build, Lovable used a total credits across nine prompts, which comes out to about $7.73. Breaking it down, most features stayed within a very reasonable range, roughly around $0.58 to $1.63, and even the more complex Stripe integration only needed three prompts and cost $1.70 while still working properly in the end. And that's really the key difference here. Everything just worked, [music] and it didn't take a lot of retries to get there. And with a pro plan giving us up to 150 credits a month, this entire app only used around 30.9 credits, [music] and that leaves us a lot more room for more projects, iterations, or even multiple builds within the same plan. Overall, Lovable ends up being one of the most cost-efficient options so far, not just because it's cheaper, but because it avoids the extra cost that comes from repeated failures and unnecessary prompts. All right, so for the final app, we're testing Base 44. And just like the others, it follows the same prompts, but the main difference here is how pricing works. Instead of tokens or credits, Base 44 uses messages [music] as its unit. It starts with 909 messages, and pricing is set at $20 for every 100 messages, so we can directly track how much each step costs based on usage. For the foundation, Base 44 completes this in a single prompt with no revisions. It uses just two messages, bringing the total to 911, and that puts the cost for this step at about $0.40. And the result is surprisingly strong. It generates a full card vault dashboard with a sidebar, collection stats, recent activity, category breakdowns, and even top cards by value. It feels more complete compared to a basic layout, and it does all of that with very minimal usage. Moving into the card collecting feature, this is where actual functionality starts getting added. Base 44 handles this in a single prompt with no revisions, using just four messages, and that brings the total to 915, and the cost for this feature coming out to about $0.80. [music] The implementation is complete. Our users can add cards, they can browse through them, apply filters, and view everything in a structured gallery. The data setup also feels properly connected as well, so it's not just visual, it actually works as a system. For the amount of functionality added here, the message usage stays very low, which is a strong sign early on. Adding the viewing modal builds on top of that by improving how users interact with each card, and Base 44 completes this in one prompt with no revisions, using three messages. And that brings the total to 918, and the cost for this feature coming out to about $0.60. Everything works cleanly. Clicking a card opens up a modal with full image, its stats, and collection details, and the interaction also feels very smooth and complete. It's another efficient step here, low message usage, no retries, and a fully working feature right away. And when it comes to login, authentication, and database setup, there's actually nothing to build here because Base44 already includes this natively. So, there's no need to use a prompt or spend any messages on it at all. That is a big advantage because it removes an entire step from our process and avoids any extra cost or potential errors. It's already built into the platform and works out of the box. Adding Stripe, premium access, and AI scanning is usually one of the most difficult parts of our build so far. But, Base44 handles all of it in a single prompt with no revisions. It uses just five messages, bringing the total to 923, and the cost for this feature coming out to about a dollar. Everything is fully integrated right away. Payments work, premium access unlocked properly, and the AI scanning feature is already functional. Now, one thing that stands out here is that Base44 has native AI built in. So, there's no need to bring in your own API key or set up anything externally. And that makes the whole process just much smoother and avoids the issues we saw earlier with other platforms. And when it comes to the collection and wish list sections, these are already in place. [music] Base44 automatically includes this as part of the app structure. So, there's no need to run an extra prompt or spend any messages on it. >> [music] >> And that's another step saved, and more importantly, it keeps the total cost lower since we're not paying to build something that's, well, already built in, right? So, after the collection tools are already handled, the app still needs a way to help users look beyond their own cards, and that's where the market watch section comes in. Base44 completes this in one prompt with no revisions using just three messages, and that brings the total to 926, and the cost for this feature coming out to about 60 cents. [music] This section is fully functional. Users can browse card values, explore trends, and use that information alongside their collection and wish list. It's another efficient step here. No message usage, no retries, and everything works just right [music] away. So, to wrap everything up, the last piece of the app is the marketplace feature, which is usually the most complex part since [music] it does combine listings, user interactions, and a separate system within the app. Base44 handles all of this in a single prompt with no revisions using just three messages, and that brings the total to 929, and the cost for this feature coming out to about 60 cents again. Everything is built in one go. Users can create listings, they can browse other listings, they can apply filters, and they can interact with each other all without needing any additional fixes or prompts. So, for something this complex, the efficiency here really stands out. It keeps the cost extremely low while still delivering a complete and working feature on the first attempt. Base44 just stands out here immediately once you look at the full build. Across the entire process, it completes the app using just 20 messages, which comes out to around $4, and that's with every feature already fully working. It also only takes just six prompts in total, which is the lowest so far. A big reason for that is how much is already built into the platform. Authentication and database setup are handled natively, so there is no need to spend prompts on them. And the same goes for parts like collection and wish list, which are already structured into the app from the very start. >> [music] >> The Stripe integration, the premium system, and AI scanning are also done in a single prompt, and everything just works without needing any revisions. That is a huge difference compared to the other platforms, where this step either fails or becomes the most expensive part. So, overall, Base44 is extremely efficient. It avoids unnecessary prompts, reduces retries to zero, and delivers working features consistently. And that combination of built-in functionality and low message usage makes a huge impact on both cost and development time. Base44 approaches pricing in a much more straightforward way compared to the other platforms, too. There's a free plan, of course, with 25 messages per month, which is enough to test things out. The starter plan is $20 a month for 100 messages, and then it scales up with higher tiers, offering more messages at better value. All plans include unlimited apps, which is useful if you're working on multiple projects. What stands out here is how predictable everything is. There are no daily limits or confusing conversions, just a fixed number of messages per month, which makes it easier to plan ahead. Now, for the build, Base44 only used 20 messages across six prompts, and that brings the total development cost to just $4. Breaking it down, most features stayed extremely low, around 40 cents to a dollar, [music] and several key parts like authentication and database setup and collection management didn't even require any prompts at all. And that's really where the difference shows. A lot of the functionality is already built into the platform, so you're not spending messages on things that should already be there. So, with the starter plan giving us 100 messages per month, this entire app only used about 20% of that total allocation, and that leaves plenty of room for more builds, iterations, or even multiple projects within the same plan. Overall, Base44 ends up being the most cost-efficient option by a wide margin. It keeps costs low, avoids unnecessary prompts, and delivers working features consistently without needing retries. So, after testing everything, the difference really comes down to efficiency. Some platforms are good at first, but once you start adding real features, then the cost and reliability starts to matter a lot more. And out of all of them, Base44 clearly gives us the best value overall. And if you want to try them out yourself, I added a special link in the description down below. I'll see you at the next one, and thank you for investing your time with me today.