Paying $80 to $100+ a month for ChatGPT, Claude, Gemini, Midjourney, and a video tool is the new normal, and it adds up fast. In this video I test ChatLLM by Abacus AI, a platform that puts every major AI model (GPT, Claude Fable 5, Gemini, Grok, Kimi, plus image and video models) in one place for $10 a month. I stress test it across text, image, and video generation, try the RouteLLM auto-routing feature, and use the agent to build a full presentation and a working web app from scratch. Then I give you my honest take on the downsides and who this actually makes sense for.
CHAPTERS: 0:00 The Real Cost of AI Subscriptions in 2026 0:37 In This Video 1:04 Why One AI Model Is Never Enough 1:58 What is ChatLLM by Abacus AI? 2:28 Every Major AI Model in One Platform 2:52 ChatLLM Pricing ($7 First Month) 3:11 Inside the ChatLLM Platform: Switching Models 3:46 Stress Test: Same Prompt Across Claude, GPT & Grok 4:32 Comparing AI Model Responses Side by Side 5:45 How to Distribute Tasks Between AI Models 6:27 RouteLLM: Automatic AI Model Routing 6:56 RouteLLM Coding vs Copywriting Tasks 7:26 AI Image Generation Test (Nano Banana Pro vs GPT Image 2) 8:42 AI Video Generation in ChatLLM 9:27 ChatLLM Agents Explained 9:44 Demo: AI Agent Builds a Full Presentation 10:56 Publishing & Hosting Your Deck Online 11:14 Demo: Building a Web App With the AI Agent 12:06 Deploying Your App to a Custom Domain 12:24 ChatLLM for Teams & Enterprise 12:53 Connectors: Slack, Gmail & App Integrations 13:18 ChatLLM Cost vs Paying for Each AI Tool 13:54 The Downsides of ChatLLM 14:32 Credit Limits & Who Shouldn't Use It 15:01 Who ChatLLM Is Actually Best For 15:27 Final Verdict: Cut Your AI Costs 10x
------------------- Clips channel @jonlawclips
I do NOT sell any products of my own. Watch out for scams.
So these days, to properly use all the AI models,
you're spending 20 bucks a month for Chex UPT,
20 for Claude, 20 for Gemini, 20 for Mid Journey
or some kind of image tool,
another 50 bucks for a video tool.
Basically, you're paying at least $100 a month,
if not a lot more than that,
just to access all the top frontier AI tools.
And the reason you have to do that fundamentally
is there's no clear winner, right?
There's no one platform or model or company
that's better at everything
versus everything else out there.
And so because Claude might be better at coding,
but Chex UPT is better for this,
but another model is better for images,
then you have to pay for all of these simultaneously.
But in this video, we're going to cover a tool
that uses a different approach,
where it gives you every single one of the top models
on top of image models, video models, everything you need,
all for $10 a month.
So we're gonna take a few minutes to explore that solution
and then stress test it across the verticals
that will matter to you
when you go and actually use this thing in practice
to cut your costs down
and access all of the leading models from one place.
That's gonna be the video.
It's gonna be a fun one.
Let's go.
Now, again, just for context here on kind of the problem,
number one is, of course,
all the leading companies are going to charge
for access to their products.
So that's fine as step one, right?
But that is something you're going to have to pay for.
But then step two here is what you should be using,
aka what is best intelligence or capabilities-wise,
again, changes all the time.
That used to be month by month.
Now it's more like week by week in some cases.
And then even within one model
being on top of the leaderboards more generally.
The third component here is that each task can vary a lot
and new models, new ways of doing things
are coming out all the time.
And most people just default
to whatever they're paying for.
That's fine, depending on what that default is.
But the better way to do it
if you want to maximize performance
is just actually switching between all of them.
But to do that, again, you have to pay for all of that.
So that in mind, the solution we're looking at
is a very popular product.
You may have heard of it called ChatLLM.
That's made by a company called Abacus AI.
So we're going to take a look at ChatLLM.
They are sponsoring this video.
This will be my unfiltered, honest take on the product.
And so we're going to walk through these demos
and then I'll give you my overall take
on the strengths and the weaknesses.
And overall, my take on how it fits or feels
as a solution to proliferation in the AI model
and consumer app space.
And the core idea of ChatLLM is one platform
with every major AI model inside of it.
So all your leading ChatGPT, your Claude,
your DeepSeek, your Gemini,
all those models are available for use within one platform.
And then whenever a new model drops,
they add it within a few days of the launch.
So we're going to look at that.
They also have a bunch of agentic features,
so agents that can go do things, build things for you.
So we're going to jump into my computer.
If you want to follow along while you're watching this,
you can do so at this link on the screen
or in the description.
It's $7 for the first month,
so you can sign up and follow along with this video
to walk through the demos and feel out the platform as well.
And then in mind, let's go.
So here we are inside the ChatLLM platform
under Abacus AI.
Now we're going to walk through
a few of the different features here.
The central one, like I've said,
is the fact that right here, right now,
we're talking to ChatGPT 5.6 sole.
And if you want to change that,
here are all the different models that we can use.
You want to use your Fable 5 from Claude, you can do that.
Kami3 has made a massive splash recently.
You can use that too.
All the leading models, open source, closed source,
whatever you want are here, as well as our image models.
So our Nano Banana,
and then your video models if you want those as well.
So to illustrate that concept
and start our stress test of the platform,
I want to give it this prompt.
So I run a YouTube channel about AI tools.
Write me a 30-day content strategy
that balances views with sponsor revenue
and be specific about video formats.
So just a general prompt
so we can get an idea of how the AIs think.
And then let's go ahead and duplicate this tab.
So we're opening the same tab,
excuse me, the same page in three different tabs.
We're going to give Fable this prompt.
Let's go ahead and let that run.
You can see the UI UX is the same as the chat box over here.
Let's go to GPT 5.6.
And then over here, let's jump down to Grok 4.5.
So for the Claude Fable response,
we can see the weekly cadence.
So repeat four times a week with new topics.
We're going to have short, trend and news,
a deep tutorial, and then a listicle or comparison.
We have our 30-day arc.
Now over here,
we have something that looks a little bit different.
We have our three long-form videos per week,
four shorts per week.
We can see something roughly similar,
but I think Fable hits the formatting a little bit better
with this cadence table.
And if we jump into Grok,
we don't even have a table of any sort.
We just have a very long text response here
that really just keeps going on.
So it tried,
but I think the formatting on the other two
is definitely better.
You can get an idea here
of what just using the platform is like,
but also critically,
the responses are not the same.
If you're using Fable versus a GPT model
versus Grok versus Kimi or Muse
or any of these other models,
the responses will be different.
So if you're using just one of them,
you're just going to get that type of response.
And if you aren't using the best models,
you are probably going to get a worse response.
All of these are relatively frontier.
But if we jump into something that's a bit older
and we give it the same prompt,
then we are going to get
just a different quality of response.
Now, a big subject in the world of AI recently
is being how you should distribute tasks between models.
So if you have some tasks that are less compute intensive,
require less intelligence,
maybe you don't give that to a frontier model.
You give it to an older model
and you only give these closed source frontier models,
the tasks which you really,
really want the best performance from.
And a platform like this can kind of do that natively,
because it has all the different models
natively embedded in it.
So let's jump back to the homepage here.
And then instead of choosing a model here,
I'm going to go to the top option that's called route LLM.
And what this does,
it's kind of like an agent or a brain
that based on your prompt,
it gives the task to a certain model.
So for example, here's a task,
it's write a Python script that renames every file
in a folder based on its creation date.
So let's go ahead and let that run.
And we can see that it routed it to CloudSonic 4.6.
So a very good coding model that also is cost effective.
And then it went ahead and it's giving us this file.
But then if we jump back into the homepage
and we give it this prompt,
so write a punchy 60 second YouTube ad script
for a coffee brand,
let's see what model that goes to.
And that goes to Gemini 3 Flash.
So a different skillset, worse at coding,
but better for this use case,
and also really, really cheap.
So in terms of efficacy of usage, right,
your spend, something like that is useful across the board,
definitely useful inside chat LLM.
So that's what we're looking at on the text side.
Now let's give the image side of the platform a go.
So we can jump into image or video here.
We can also just leave it in the default chat box here.
And then we can just select a model
or let route LLM do it for us that does images.
So for example, let's go Nano Banana Pro
and then give it this prompt.
A YouTube thumbnail style image of a person at a desk
surrounded by AI logos, basically a thumbnail ask image.
And then let's head back here, go to image,
just so you can get an idea of what that looks like.
We'll select GPT image two.
So same prompt, but a different model.
And let's see what that means.
So here's the resulting image from Nano Banana Pro.
We have a thumbnail AI revolution spelled correctly.
So kudos to that with our little floating logos here.
And let's see the GPT two image.
Hopefully this one's better, we'll see.
Okay, and here we can see a image
that has the actual models and the real logos,
also more text.
So if we compare and contrast these two models,
this is what I actually suggest you to period
with your image generation across the board,
use different models, see what's better
and also think about cost as part of that equation.
So that's image, we can go for video as well.
So for example, let's click use as start frame
and then do a slow pan in on this man.
And then we can go auto or we can choose a model.
I'll just leave it as auto for video.
That's all of the leading models right there.
And let's see what this makes.
Okay, here's our video.
So we have a fun little pan in here.
I don't know about the audio here,
but if you want that as B-roll, that would be pretty cool.
And so there we go, that's video.
So what we've explored so far
are the various generation features.
So text by just chatting to these different LLMs
and then your image and video models as well.
Now, something else you can do is go to the agent feature.
The agent is basically something
that using all these different models
and routing between them can do things on your behalf, right?
Can build these soup to nuts products for you.
So let's give it this prompt,
create a professional presentation.
So a PowerPoint or a deck on the state
of AI video generation in 2026,
including the models, what they cost
and where the industry is heading.
We'll just leave this on the default setting
and let's let it run.
Okay, it started by asking any questions.
I'll just say, go ahead and we'll choose the best details.
Okay, here we can see that it is searching.
So it's doing a bunch of research on this
on all these different websites
and then it'll start building the deck.
Okay, we can see it's built a full report here basically
and then that's what it's now turning into the presentation.
Okay, here we can see the finished deck.
Looks good, we have 14 slides.
Numbers look like they are accurate.
This is all researched well.
We have a good looking organization there.
We have a little graph here displaying the cost.
That's actually really useful.
So that looks good.
Yep, overall, that is what we wanted
and we can go ahead and export that
to our other slide platforms
or we can go ahead and straight up publish
so we can access that anywhere on the internet like so.
This is actually a little mini feature
that I don't really see that often.
I like it where you can just host a presentation
on a kind of public internet link that anyone can access.
I think more people should do that
but anyway, here you go.
That is our presentation.
Good, solid build right there.
But let's even go farther with the agent feature, right?
We just made a deck.
Let's build a full on app.
So build me a web app called AI Tool Tracker
where you can browse a database of AI tools
and we want a good looking design there.
Now also right here, we can choose how much effort we want.
In this case, let's go for some pretty high effort.
Let's go for max effort.
We can add skills.
In this case, we don't really need to
and let's let that build get started.
Okay, here we can see the app has been completed.
This is actually looking really good.
It added all of the tools for us.
We can add more if we want.
That's definitely a good start
and we can go between our paid, premium and free.
We can submit a tool.
That looks good.
That all works.
We can go back to browse.
So yeah, there we go.
That is a full web app
and all the hosting, the backend is completely built in
so we can go ahead and click deploy
and then we can access that on the internet
under a custom domain
or if we wanna just deploy it quickly, a managed domain.
We can also click on the app management console right there
and we can see all the different deployments of the apps
and we can do mobile as well as web.
So pretty cool set of features there.
A few other quick notes as well.
If this is something used in a business context,
you're going to want to add team members
so you can see all your team members.
For example, here.
So you can manage all your models from one place.
You can cut costs at the enterprise level.
That's the whole token maxing trend
that is slowly dying out
due to just the insane costs associated
with not switching between your models
depending on task difficulty
and then all the different conversations
of people in that team are public between everyone.
And then second, there are connectors.
So if we click under connectors right here,
we can add things like your Slack, your teams,
basically your whole enterprise
and personal suite of productivity tools.
So if you want to pull data, right,
if you want to let your agents do things,
take actions on your behalf in your Gmail,
your YouTube, your social media,
then you can do it by adding those as connectors here.
So as far as cost again,
if we think through the price
of each of the individual model platforms,
it's $20 a month and up as a general rule of thumb
for your Plod, your ChatGPT, your Gemini Suite,
Grok, et cetera, and then the net cost.
And I think the big advantage of a platform like this
is it's $10 per user per month.
So that is the cost differential there.
And that overall is, I think,
everything you need to know about the platform.
We took a look at pretty much all the features
that I think differentiate it
versus using those models individually.
And now let's talk through some of the downsides
or the other side of using a platform like ChatLLM.
Overall, those are the demos.
Take what you will from there as far as performance goes.
Now, my overall take is if you are a power user
of one native ecosystem, like you love Plod, you love GPT,
those native ecosystems are going to have
some of the more edge features and new additions first.
So if you love one of those and you're not price-conscious,
you don't care about spending some extra money,
then using just those native ecosystems
is going to be the better solution for you.
So that's the first thing to keep in mind.
The second is that the subscription cost on ChatLLM
is not unlimited everything.
If you want the more credit-intensive edge features,
like video generation, that's going to burn through
that kind of credit equivalent for the subscription
faster than just something like ChatWill.
So if you have a more enterprise-style business use case
where you want to do something bulk,
like you want to generate 300 video clips a month,
this is not going to be a fit for you as well.
So for people in those two categories,
it might be fun as a kind of consumer thing
just to have access to try out all the models,
but it's not going to be the ideal
kind of solution here for you.
So what I think it's best for is
if you pay for all of these tools,
but you just want a way to use the different tools out there
that are the best, especially for lighter searches,
for chats, this is going to be perfect for you
because you're going to go from spending $100 a month
to $10 a month.
The math on that, just maths.
And so I think that is the kind of use case
presented here by ChatLLM.
And so there you go.
That is every model in one platform
in a very feasible path to cut your AI costs by 10x.
Now, again, if you want to support the channel
and try this out, you can get your first month for $7
at this link on the screen or in the description.
And what I want you to do is drop in the comments
the amount that you're spending per month on AI tools.
I think that'll be fun.
I'll add what my monthly subscription costs are.
And so I appreciate y'all being here
and I will see you in the next one.
Cheers.