The argument in one line.
The 2026 AI landscape splits into four essential paid models (ChatGPT, Claude, Gemini, Grok) plus dozens of free open-source alternatives, with specialized tools for video, images, and agents that let anyone automate entire business tasks.
Read if. Skip if.
READ IF YOU ARE…
- You're new to AI and want a 25-minute map of which tools exist, what they do, and which ones actually matter for your workflow.
- A creator or business owner who uses AI daily but hasn't evaluated the full landscape since last year and wants a current state-of-the-stack refresh.
- You're considering paid AI subscriptions and want concrete comparisons of pricing, capabilities, and use cases before committing money.
SKIP IF…
- You're building AI products or fine-tuning models — this is consumer-tier tool comparison, not technical architecture or training methodology.
- You need deep dives on any single model — this trades depth for breadth and moves fast through each tool without implementation detail.
The full version, fast.
The AI stack in 2026 has consolidated around four paid chat models, plus open-source contenders, specialized media generators, and agents that execute work autonomously. ChatGPT wins on memory and image generation, Claude leads writing and coding, Gemini delivers the best value and multimodal analysis, and Grok offers real-time data and the largest context window at the lowest price. Open-source options like DeepSeek, Llama 4, Qwen, and Mistral now run locally and close the gap for everyday tasks. Video tools (Seedance, Veo 3.1, Kling, Runway) and image tools (GPT Image 2, Midjourney v8, Flux 2) cover production needs. Pick the paid model matching your dominant workflow, then deploy agents through Claude Code or self-hosted OpenHands to compound output.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:58
01 · Cold open + promise
Names every AI surface in 7 seconds, lands the 'while you sleep' match-cut, promises 'you'll learn something new'.
00:58 – 03:29
02 · ChatGPT
GPT-5.5, free vs Plus vs Pro pricing, memory feature, image gen, multistep tool use, voice mode, 1M context window.
03:29 – 06:10
03 · Claude
Opus 4.7 as the best writer + best for coding, projects feature, 1M context, self-verification, Claude Code agent mention.
06:10 – 08:10
04 · Gemini
3.1 Pro, generous free tier, Google Workspace integration, multimodal video/audio strength, cheapest frontier API at $2/M tokens.
08:10 – 10:40
05 · Grok
Grok 4.3 inside X, $4/mo with X Premium, real-time tweet data, downloadable PDFs/PPTs/sheets, 2M context, cheapest API at $0.20/M.
10:40 – 11:19
06 · Summary slide
One-line positioning per model - best all-around / best writer / best value / cheapest.
11:19 – 14:20
07 · Sponsor: Bluehost All Access
$20/mo gateway pitched as the bridge between 'paid four' and 'free open source' sections. ~3 min walkthrough with demo of model-switching in one chat UI.
14:20 – 16:36
08 · Open-source models
DeepSeek v3.2, Llama 4 Scout (10M context), Qwen 3.5 / Kimi K2, Mistral. Ollama as the on-device runner. VRAM requirements.
16:36 – 18:21
09 · AI for video
Seedance 2.0 (unified audio+video), Veo 3.1, Kling 3.0, Runway Gen 4.5. Sora 'shutting down' claim. Higgsfield as aggregator.
18:21 – 19:50
10 · AI for images
GPT Image 2.0, Midjourney v8 Alpha, Flux 2, Imagen, Ideogram. Use-case routing rules: GPT for text, MJ for art, Flux for photoreal.
19:50 – 22:30
11 · AI agents
Claude Code / Claude Cowork explained (folder access, multi-agent, MCP). OpenClaw as Telegram-based VPS-hosted personal assistant - car-negotiation anecdote.
22:30 – 23:30
12 · Sponsor: Bluehost VPS
Pitch to host your own OpenClaw on a VPS for ~$2/month.
23:30 – 24:35
13 · Close + CTA
Use AI or fall behind. Like, subscribe, see you next video.
Lines worth screenshotting.
- ChatGPT is the best all-in-one model for beginners because it does everything adequately, while Claude wins on writing quality and long-document analysis.
- Claude Opus 4.7's self-verification feature — writing and running tests before delivering an answer — is directly valuable for accuracy-sensitive tasks.
- A 1,000,000 token context window across both ChatGPT and Claude effectively eliminates the document size constraint that previously forced chunking strategies.
- GPT image generation now supports 2K resolution and 8 images from a single prompt, making in-chat image production viable for production-quality assets.
- The AI model landscape in 2026 spans chat, open-source, video, image, and agent categories — no single model dominates all five simultaneously.
- Claude Cowork is the lower-complexity entry to agentic AI; Claude Code offers the same power with a steeper learning curve oriented toward developers.
- Open-source models matter primarily for users who need data privacy, cost control, or local deployment — not as general-purpose alternatives to frontier models.
- AI video and AI agent categories are evolving faster than chat models in 2026, meaning evaluations that were accurate six months ago are likely stale.
- The best use case for each model follows from its structural strength: ChatGPT for ideation and memory, Claude for writing and code, Gemini for multimodal and long video.
- A bundled discount service offering access to multiple paid AI models at 75% off is a product category that only exists because model switching costs have become real.
- Claude's dominance in developer tooling (Cursor, Windsurf, GitHub Copilot) means it touches far more code than its direct subscription numbers suggest.
- GPT 5.5 releasing six weeks after 5.4 signals that model iteration cycles have compressed from months to weeks, making version tracking a real cognitive overhead for users.
Steal the promise-stack hook.
Wide-net longform playbook
When the topic has many named brands, list them all in the first 7 seconds - every brand name is a separate search-intent magnet.
- Open with a list of every famous brand in the category you're covering. Each name is a search-intent grab - viewers who came for ChatGPT also stay for Claude/Gemini/Grok.
- Pair the dream-outcome promise with a comedic match-cut B-roll (he says 'while you sleep' and cuts to someone asleep). Costs nothing, buys retention.
- Build a one-screen summary slide at the structural midpoint (10:40). One line per option. This is the frame people screenshot and re-share.
- Place the sponsor at a structural pivot, not an arbitrary 6-minute mark. Frame it as the bridge between sections, not an interruption.
- End every sub-section with 'I personally use this for ___' - that's the line viewers remember, not the spec table.
- Add a re-hook every 6 minutes ('there's this whole other world of free open-source AI'). Each re-hook is a fresh promise that resets attention.
- For Mod Boss specifically: the 'rent vs own' frame is built into his open-source segment - riff on that for the $6 Stack pitch.
Terms worth knowing.
- open-source AI model
- An AI model whose weights, architecture, or code are publicly released — allowing anyone to download, run, modify, or fine-tune the model without a subscription, and without sending data to an external company's servers.
- frontier model
- A state-of-the-art AI model representing the current performance ceiling of publicly available systems — typically the most capable and most expensive to run at any given point in time.
- multimodal
- The ability of an AI model to understand and generate multiple types of data — text, images, audio, and video — in a single interaction, rather than being limited to one format.
- token (AI pricing)
- The basic unit of text processed by an AI model — roughly equivalent to a fraction of a word — used as the billing unit for API access, with pricing typically expressed as cost per million tokens.
- Ollama
- A free, open-source application that allows users to download and run large language models locally on their own computer — enabling private, offline AI use without sending data to external servers.
- VRAM
- Video RAM — the dedicated memory on a graphics processing unit (GPU) used to run computationally intensive tasks — the primary hardware constraint for running large AI models locally, with most models requiring between 8 and 24 gigabytes.
- VPS (virtual private server)
- A rented, always-on cloud-based computing environment that functions like a dedicated server — used here to host AI agents that need to run continuously without depending on a personal computer remaining powered on.
- fine-tuned model
- A version of a base AI model that has been further trained on a specific dataset to improve performance for a particular task or domain — common in the open-source community where developers customize models for specialized use cases.
- SVG (scalable vector graphic)
- A file format for two-dimensional graphics defined by mathematical equations rather than pixels — enabling infinite scaling without quality loss, and capable of encoding interactive or animated elements for web display.
- lip sync (AI video)
- The automated alignment of a video character's mouth movements to a provided audio track — a key quality metric for AI-generated talking-head video, where poor sync creates an uncanny mismatch between speech and visible articulation.
- benchmark (AI)
- A standardized test or dataset used to measure and compare the performance of AI models on specific tasks — such as graduate-level science questions, coding challenges, or reasoning puzzles — enabling objective comparison across different systems.
- MCP (Model Context Protocol)
- A standardized interface that allows AI agents to connect to and control external applications and services — enabling agents to read email, manage calendars, browse the web, and take actions across a user's digital tools from a single conversation.
- OpenClaw (Open WebUI/agent platform)
- An open-source AI agent platform that allows users to run a persistent, self-hosted AI assistant — connectable to email, calendars, and external services — typically run on a VPS so it operates continuously without relying on a personal device.
Things they pointed at.
05:26toolCursor
05:26toolWindsurf
05:28toolGitHub Copilot
05:30toolClaude Code
05:50toolClaude Cowork
11:19productBluehost All Access
15:45productMeta Llama 4 Scout
16:10productQwen 3.5 / Kimi K2
16:28productMistral
16:36toolOllama
17:00productSeedance 2.0 (ByteDance)
17:20toolHiggsfield
17:35productVeo 3.1 (Google)
17:55productKling 3.0
18:05productRunway Gen 4.5
18:40productGPT Image 2.0
19:05productMidjourney v8 Alpha
19:20productFlux 2 (Black Forest Labs)
20:00productOpenClaw
22:30productBluehost VPS
Lines you could clip.
00:05
“AI agents that can literally do work for you while you sleep.”
03:55
“If you give the same exact prompt to ChatGPT and to Claude, you can actually see a very, very big difference. The output just feels more natural, feels more thoughtful, more human, less robotic.”
05:20
“It has a 1,000,000 token context window. So you can basically upload an entire book or a super long contract or a very long report.”
11:30
“You actually get four paid versions of ChatGPT, Claude, Gemini, and Grok in one single plan.”
23:50
“If you're not taking advantage of AI, other people are, and they're gonna move way faster than you.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogystory
00:00Hey, guys. My name is Charlie, and in this video, I'm going to explain every single AI model you need to know about right now. And I mean all of them.
00:06ChatGPT, Claude, Gemini, Grok, the free open source ones, AI for video, AI for images, and even AI agents that can literally do work for you while you sleep. Obviously, the AI space is moving super, super fast, and most people have no idea which models are actually worth using and which ones are a waste of money.
00:24So what I'm gonna do is breaking down exactly what each one does, show you real examples of each one in action, and tell you which ones I actually use and recommend. Now whether you are a complete beginner or you've been using AI for a while, I promise you're gonna learn something new in this video. So let's get started.
00:38So let's start with the four main AI models that most people are going to be choosing between. And if want the paid versions of all these, I'll show you a that actually gets you access to all of them for about 75% off. But more on that later in this video.
00:51Okay. First on the list is, of course, ChatGPT from OpenAI. Anyone that's ever used AI before has probably used ChatGPT, and that's because it's the most popular one.
01:00So they actually just released GPT 5.5, and it's the first fully retrained base model since GPT 4.5, which was quite a while ago.
01:08Now depending on when you're watching this video, this may be relevant or this may not be relevant. Just know that OpenAI is always releasing new versions of GPT, and each one gets significantly better. These are also moving a lot faster right now.
01:19So GPT 5.5 was released six weeks after GPT 5.4. So just keep that in mind.
01:25Right? Things are moving really fast. Now there's a free tier of ChatGPT.
01:29You can chat with us. You can generate images. You can browse the web.
01:32You can upload files. It's honestly really solid for, you know, not having to pay anything. There's also the paid plan, is ChatGPT Plus.
01:39This is $20 per month. Super, super worth it. And with this, you get full access to the latest models, so that being 5.5 right now.
01:45You get a lot more usage. You get priority access, advanced voice mode. And there's also the pro and enterprise tiers, which most of you guys won't need, and those are a lot more expensive.
01:54Now personally, I love ChatGPT because in my opinion, it has the best memory features. So it remembers things across all of my conversations, like my preferences, my business, projects that I'm working on, and so you don't have to re explain everything every single time.
02:06What I like to use ChatGPT for is a lot of brainstorming and a lot of ideation and a lot of, like, business developments and bouncing ideas off of someone. And by someone, I mean ChatGPT. It's a really good mentor.
02:17It understands me really well. It gives really solid advice. And, yeah, that's why I'll use it for most basic tasks as well as tasks where I'm asking it questions about life, about business, about anything.
02:28It can also generate images right inside the chat. So they actually retired Dolly, and they built ImageGen directly into the model. So, yeah, gbt image, it's honestly one of the best for text on images.
02:38Like, if you need a logo or a social media graphic with readable text on the actual photo, it can actually do that really well. GPT image two point o just released. It actually supports two k resolution and can generate up to eight images from a single prompt.
02:52You guys should try this out. It's really, really good. GPT 5.5 is also a lot better at complicated multistep tasks, so you can actually give it pretty messy requests like research these five competitors, build a comparison spreadsheet, and write me a summary.
03:06And then it's going to plan it out. It's gonna use the tools. It's gonna check its own work, and then it's gonna keep going until it's done.
03:11There's also voice mode, so you can actually talk to it just like with a person. And in my opinion, I think it's the most, like, natural voice AI out there right now. Now GPT does have a 1,000,000 token context window, which is solid.
03:23And, yeah, if you are brand new to AI and you just want one single tool, then it's probably gonna be ChatGPT. It does a little bit of everything pretty well, and that's why I think it's the best, like, all in one overall one. Okay.
03:34Next, let's talk about Claude, which is from Anthropic. I've been using Claude a ton recently. I think it actually is my most used AI model.
03:41And currently, their flagship model is Claude Opus 4.7, which just came out April 16. You get access to this if you have the paid plan for Claude, is, like, $20 per month.
03:51There's also a free tier, so you do get access to Sonnet for free with daily limits. And with that, you can chat, you can upload documents, you can analyze files. It's very similar to ChatGPT.
04:01Now like I mentioned already with the paid Claude Pro plan, which is $20 a month, you get full access to the latest model, which is currently OPUS 4.7. You get way more usage. You also get access to Claude's project feature where you can upload a ton of documents and remembers everything.
04:15In my opinion, it out there when it comes to writing. If you give the same exact prompt to, let's say, ChatGPT and to Claude, you can actually see a very, very big difference. The output just feels more natural, feels like more thoughtful, more human, less robotic.
04:32And if you're writing blog posts, scripts, or emails, anything where the quality of the writing matters a lot, then I think Claude is the best one to use. Claude's also really good at handling really, really long documents.
04:43So it has a 1,000,000 token context window. So you can basically upload an entire book or a super long contract or a very long report, and you can ask it questions about the whole thing. There's also the projects feature, and this is super underrated in my opinion.
04:56You can actually upload a bunch of reference documents into a project, and will use all of them as context for every single conversation. So for example, if you have, like, brand guidelines or product documents or style guides, throw them all in there into a project, and then Claude just knows your business. Now with OPUS 4.7, it does have high res image support now.
05:16So up to, I believe, 2,576 pixels. It can read screenshots.
05:21It can analyze charts. It can even do pixel level data extraction from graphs. Also, a really cool thing about 4.7 is it does self verification.
05:29So it proactively writes, tests and checks its own work before giving you the answer. So that's a really good thing when it comes to accuracy. If you're trying to do things that require very accurate information or data, then Claude is really good for that.
05:41Claude also, of course, dominates the developer tool space. So Claude is the AI model between, like, Cursor, Windsurf, GitHub Copilot, and also Claude Code.
05:50Now personally, I'm running a bunch of different Claude agents within my own business. If you wanna have an AI that can actually perform tasks, I highly recommend looking at Claude Cowork or Claude Code. Cowork is a bit easier to use in my opinion, but can absolutely, like, create files and edit files on your computer.
06:05You can build out different agents and schedule tasks, cron jobs. Claude Code can basically do everything that Claude Cowork can do. It's just a little bit more complex looking, and so I'd say that's a bit more for advanced people.
06:16When it comes to a lot of my business output and actual, like, actions that are happening within the business as well as my personal life, I tend to like to run them through Claude. For example, my email assistant, my research assistant. I get a lot of help writing scripts with Cloud as well.
06:29Overall, it's an insanely good AI model right now, and the $20 for the paid plan is absolutely worth it. Next, let's talk about Gemini from Google. And, honestly, Google's had a really big comeback here.
06:40They fell back behind for quite a bit, but now they have Gemini 3.1 pro, which is their paid plan. And there's also the free tier. So Gemini is free to use on gemini.google.com.
06:51It's also built into Google Workspace. So if you use Gmail or Google Docs or Google Sheets, then it's perfect. In my opinion, the free tier is actually really generous.
07:00You can do a lot with it, and it performs really well. Now the paid plan, is Google AI Pro, it's currently $19 per month.
07:07And when you pay that, it also comes with other stuff like Google One, which gives you a bunch of storage. So it's not just like the AI model with this plan. Like I said earlier, if you guys use Gmail or Docs or Sheets, Gemini is built right in.
07:19So you can have it draft your emails. You can have it summarize threads. I can create documents.
07:23It can build formulas and sheets. And the great thing is that you don't need to actually leave the apps that you're using. It's built in.
07:30A lot of the time if I'm in Gmail, then I'll use Gemini to actually help me draft a reply. Or if I'm actually creating stuff within a Google Doc, it can help me write things, revise things, and edit things. Now in my opinion, Gemini is probably the best at understanding different types of media.
07:44So you can, for example, upload an hour of video or tons of hours of audio, and it can actually analyze the whole thing. How this works out in real life is let's say you have a recording of a super long meeting, you can record it and ask what were the action items, and it will tell you. Gemini 3.1 Pro actually leads multiple benchmarks as well, so it scores extremely well on graduate level science tests, and the reasoning performance is also really good.
08:07Gemini can also generate animated SVGs and three d code right from a text prompt, so most of the AI models are not good at that. For example, you can describe a chart or a diagram, and it's going to create an interactive animated version.
08:21It's also the cheapest frontier model for developers. So currently, it costs $2 per million input tokens versus $5 for GPT 5.5 and $5 for Claude Opus. And so, yeah, if you're already in the Google ecosystem ecosystem or you need to analyze a lot of different types of media, I think Gemini is one of the best for that.
08:40Plus the pricing is super competitive. Okay. Next, let's talk about Grok.
08:44This is made by Elon Musk's company x a I, and it actually lives inside of x, which is, of course, formerly Twitter. And as of right now, when I'm filming this video, we're currently running Grok four with the latest ones actually adding a ton of crazy features. Now there is a free tier, so basic Grok access is free on X.
09:02It's a bit more limited, but you can still chat with it. And then the paid one, this comes with x premium. It's $4 per month as of right now.
09:09This gives you full Grok for access, which makes it the cheapest paid option of the four, and there's also super Grok for $30 per month with the heavy model. Now one thing that makes Grok super unique is that it pulls from x in real time. So if you ask it about breaking news or trending topics or what people are saying about this company right now, it has a live data and the other models don't have that.
09:30For example, I can ask Grock about something trending right now, and you can see it pulls real time tweets and data. With the latest version of Grock, which is 4.3, it can now generate downloadable PDFs, fully populated spreadsheets, and even PowerPoint decks right from your conversation. So there is a lot of business use for Grok two.
09:48I think a lot of people think of Grok as just, like, built in with x, but there's actually a lot of business stuff you can generate with it as well. 4.3 also added native video input, so you can share a video clip and have a conversation about what's in it. Now this is a bit more advanced, but 4.2 also introduced a four agent system where different AI agents handle different parts of your request at the same time.
10:09So it's basically cross verifying each other's work. It's like using multiple models at the same time. Now Grok does have a 2,000,000 token context window, which is double what ChatGeeBoutique and Claude have.
10:19And then API pricing starts at 20¢ per million tokens, which is by far the cheapest for developers. Overall, like here compared to the other AI models, but it's really good. It's really fast.
10:31It's the cheapest paid option. You get so much real time data as a huge context window, and it's definitely worth trying. Okay.
10:38So just to summarize, right, we have ChatGPT. This is the best all around one, in my opinion, with best image generation and the best ecosystem. And you have Claude with Opus right now, and that's the best writer.
10:48It's the best for coding, and it's the most accurate for documents. Then you have Gemini, which is arguably the best value. It's also the best multimodal video and audio one, and it has that native Google integration.
10:59And then we have Grok. Right? This is the cheapest.
11:01It has real time data, and has the biggest context window. Honestly, you cannot go wrong with any of them. It really doesn't matter too much which one you use, but it really comes down to what you need.
11:10Right? There are definitely some advantages to having certain models, especially the paid ones. And so without doubt, like, my AI subscriptions, they are the best money I spent.
11:21It gives me a better model. It's a lot better than the free ones, and that's why I highly recommend if you wanna take AI seriously and actually use it a lot, get the paid versions of these AI models. So now I know a lot of you guys are probably thinking, okay.
11:33Cool. There's these four great models. I want to try the paid plans, but I don't wanna have four different subscriptions, and also that's pretty expensive.
11:41I'll put a link down below, but I highly recommend checking out Bluehost all access. So I just found out about this. You actually get four paid versions of ChatGPT, Claude, Gemini, and Grok in one single plan.
11:52It starts at $20 per month, and it allows you to use all of these AI models in one single place. I'll show you guys how to use this, but for example, let's say you're trying to write a document or a script, I would just switch to the cloud model and have it do that. And if I wanna do something that ChatGPT is better at, I'll switch to ChatGPT.
12:08I honestly don't know how they're doing this, but it only costs $20 per month. So if you compare it to if you buy all four, I think it's, like, 70 or $80 per month. So you're saving a ton of money, and it makes a lot of sense to get this.
12:20So, again, I'll put that link down below. It's gonna take you to this exact page. I'll click on see plans.
12:25You'll see they have two different packs. I think for most of you guys, the normal one for $20 per month is gonna be good enough. But if you want privacy, so if you are working with, like, really private information, I probably would go with the privacy plus one.
12:38It's just $5 more per month and adds another layer of security. But for most of you guys, I think this one's gonna be good, so we'll click on choose plan. You'll first want to select the duration, so monthly or one year.
12:50It's really up to you guys how long you wanna get it for. You'll select the number of users. One's gonna be good enough for most of you guys.
12:55And for example, if I go to monthly and I just have one user, it's $20. And after you create your account and put in your billing information, you just click on submit payment. Okay.
13:03So once you log in to your account, it's gonna look something like this. This is the main dashboard. You'll have all your chats over here.
13:08So this is a brand new demo account, which I just made. You can look at the different agents that you have right here. Here.
13:13But let me go back to new chat, and as you can see, here's where you're gonna just use it just like ChatGPT or Claude or whatever. This button right here, you're gonna click on this, and you can see auto route is selected.
13:23So it's going to choose which model thinks is best for your particular prompt. Of course, you do have the option to manually select. So for example, we can choose GPT, Gemini, Claude, Garak.
13:35If you want a faster response, we can choose some of these models. But let's say I wanna go with Claude Opus 4.7, I can write something like this.
13:42Write me a 500 word blog post about Claude Opus 4.7. You can also upload different files, or you can click on this to do a web search, but I'm gonna turn that off. Now we'll click on this button right here to submit, and that's gonna show you all the thinking.
13:54It's exactly just like if you were to do this within Claude. It's telling you all the different steps it took to generate this answer. And, yeah, this is that 500 word article.
14:04Of course, we can tweak it how we want. We can tell it to use certain types of wording. We can tell it to add a comparison.
14:10We can tell it to talk more about a certain topic. We can have it write in a certain tone or voice. And so now you can see this chat right here is here in our chats.
14:19We can search through all our different chats once you start building up multiple chats. You can see which model we use to generate this. So it shows right here.
14:26And, course, we can ask a follow-up. We can have it tweak different things. But for this video, I'm not gonna show you that.
14:31If you also have a team, you can actually add users right here. If you go to settings, you can control which LLMs your team has access to. Super, super cool.
14:39Right? It gives you access to all the pro models. So if you wanna compare different models and not have four different actual subscriptions, I highly recommend just using this.
14:47We've already talked about the free versions of the big four, but there's this whole other world of AI models that are completely free and also completely open source. Open source basically means the code is public. Anyone can use it, modify it, or run it on their own computer, and so you don't need a subscription, and there's no monthly fees.
15:05The cool thing about this is that it can actually run on your computer. Right? So you own it, the data stays with you, and there are a ton of them.
15:11So first is DeepSeek. We're currently on version 3.2. It's from a Chinese company.
15:16Works extremely well, and it's very strong at particularly reasoning and coding. DeepSeek v four Flash is also the current budget leader for developers.
15:25It's only 17¢ per million tokens. You can use DeepSeek for free on their website.
15:29It's just chat.deepseek.com. Next is MetaLlama four. So this is Meta's open source model, and Llama four Scout has a 10,000,000 token context window.
15:38It's free to use. There's a huge community, and there's lots of little fine tuned versions for specific tasks. Then you have Quen 3.5, which is from Alibaba.
15:46This model actually matches some of the paid models on coding benchmarks, and Kwon's Kimi k 2.6 variant actually leads the open source live ranking. Then you have Mistral, which is a French company, and this model can actually run on a single consumer GPU with 24 gigabytes of vram. Now the easiest way to actually use these free models is through web interfaces.
16:07So just going to the official websites. You can also run them locally. So for example, you can download Olama, which is a free app.
16:14You can pick a model, and you can run it on your own computer. The great thing with this is that the data never leaves your actual computer. I will say your computer does need to be at least okay or decent.
16:24You'll need about eight gigabytes of vram for some of these smaller models, but you'll want 24 gigabytes for some of the bigger ones. I will say the gap between open source models and these paid models has gotten a lot smaller. And for very basic everyday tasks, you don't need to pay for a subscription.
16:41You can simply use any of the free models we talked about before or any of these free open source ones. Okay. Now let's talk about AI for video.
16:48AI for video is getting absolutely crazy. This is where we talk about some of the more specialized AI models. And if you've been on social media, there's a chance that you guys have probably seen a video that was fully AI made and you didn't notice.
17:01It is getting better and better every single day, and soon I think it's gonna get to a point where you really cannot tell. Some of the top models for video are gonna be CDance two point o from ByteDance. This is probably the overall best value to the first model with unified audio video generation, meaning that the sound is generated with the video.
17:20It's not, like, slapped on afterwards. For example, you can generate a selfie video, and then once they flip the camera to the room that they're in, the mic is gonna sound a little bit different. You can access all these models on something like Higgs Field, which in my opinion is one of the best AI video platforms.
17:34Then you have VEO 3.1 from Google. So this is the best all around quality one.
17:39You have native four k resolution. It has, in my opinion, the best physics and, like, realism. It's also the best lip sync for talking head content.
17:46So yeah, maybe I'm using VEO 3.1 right now. Who knows? Then you have Cling three point o.
17:51This can currently generate clips in four k. You can also do multi shot sequences with consistent characters across different camera angles. You also have runway gen 4.5, so this is probably the professional's choice.
18:02Built for video editors with really precise camera controls like hand, tilt, and zoom. You might be wondering what happened to Sora from OpenAI. They're actually shutting that down, and they're just shifting their resources to their main models.
18:14Overall, if you're a content creator, this is both super exciting as well as a bit scary because you can basically generate anything now. It used to cost, like, thousands of dollars to generate really good video with AI, and now the pricing is just down so much. Now let's talk about AI for images.
18:28Images were the first big breakthrough for AI, and the models that we have for image generation now are insane. So we sort of mentioned this already, but GPT image two point o from OpenAI, it just launched and it's built right into ChatGPT. It's the best at understanding really complex prompts and text rendering.
18:45So I think it has, a 95% accuracy for text and images, which is insane. If you used to look like, let's say, six months ago at images with text inside of them, usually, it would just look like garbled text, and now it can do some pretty insane stuff.
18:58It's the first open AI image model with native reasoning, so it actually thinks about what you want before generating the image. We also have mid journey v eight alpha. So this was recently launched as well.
19:09And I'd say that if you want something that looks really designed or polished or, like, editorial quality, then the journey is probably the move. It does cost some money though, so keep that in mind. We also have flux two from Black Forest Labs.
19:21This is probably the best for photorealism, so things like skin textures and lighting and natural poses. The stuff you can make with this looks like a real photo.
19:29It's also gotten extremely fast at actually generating the image. There are a bunch of other ones like Imagen and Ideogram. Yeah.
19:35There's something to talk about. Those are the main ones. Personally, would use ChatGPT for convenience and text and images.
19:41I'd use Midjourney for art. I'd use Flux for photos or things that need to look extremely real. Yeah.
19:46It's just crazy how far we've come with AI image generation. And then let's talk about AI agents now. Right?
19:52So this is some of the more advanced stuff. So if you're just a beginner, then maybe you don't need to listen to this part. But if you wanna use AI for actually doing work, for, like, having a team of people that are constantly working for you, then this will be an important section.
20:07Instead of just, like, answering questions or giving you good ideas or generating basic things like images, AI agents actually do a bunch of different things and can actually carry out a complete task. You give it a goal, and then it's gonna figure out how to get it done. I'll say every major AI company in 2026 is focused on agents right now.
20:26It is the future of AI. And the two that you need to know are gonna be Claude Coburg slash ClaudeCode as well as OpenClaw. We've talked about Claude Coburg or Claude Code for a bit in this video.
20:35This is from Anthropic. It's honestly incredible. And, yeah, it's my most used AI agent.
20:40These AI agents actually have access to a folder on your computer so they can actually edit and build different things. You essentially just describe what you wanna build and you have a conversation with it. It's essentially vibe coding.
20:51Right? You're telling it what you wanna build. It's gonna ask you different questions.
20:54You're gonna give it more context. You're probably gonna have some fixes and stuff, and you just have a real conversation with this master coder, and it can do all the stuff for you. You can also have multiple agents working on different parts of your business or projects simultaneously.
21:08Like I said, these agents have access to your computer, so they can also open up a browser and click through different websites, fill out different forms, and actually do things visually. Right now with AI agents, you can build full stack apps from scratch, and, yeah, you can build entire systems within your business. These AI agents can connect to thousands of different apps through MCP.
21:27So for example, you can have Claude working within your Slack. You could probably go out and build an entire app or a game in one single afternoon. And it's overall Claude code and Claude Cobre.
21:36It's amazing. Next is OpenClaw. So if you're in the world of AI, you've for sure heard of OpenClaw.
21:42That's essentially why everyone was buying these Mac minis to actually run Openclaw on. Essentially, Openclaw is an open source personal AI assistant. It is free to use.
21:52You just need to host it somewhere. And like I said, a lot of people were using Mac minis. I don't think that's the best way to do it.
21:57I host mine on a VPS. This is a virtual private server. It is essentially a computer that's always on, which is in the cloud.
22:04Yeah. Overall, OpenCloud is wild. You essentially talk to it through any messaging app like Telegram.
22:09It runs twenty four seven, and it can basically connect to your entire life. For example, what you might wanna do is connect it to your email, and it's gonna send over a brief every single morning where it checks your calendar, it summarizes your emails, it pulls any news or topics you care about, and sends it all to you on Telegram.
22:26You can also talk to it within Telegram, so you can give it a certain task to do within Telegram, and it's going to do it. Think of it as like a person sitting in front of a computer. You can talk to this person twenty four seven, give them commands, they're and going to do it.
22:38For example, there was a guy who set up OpenClot to negotiate a car purchase. So it basically contacted, like, 12 different dealerships, both through email and also through messages. It tracked all the different offers in the spreadsheet.
22:49It followed up automatically and actually got him a pretty big discount. I'm not gonna talk about OpenClaw in too much detail in this video, but just know that this is one of the most popular AI agent platforms that people are using. You plug in any single AI model that we talked about in this video, and it can actually do things for you.
23:06Now like I said, I like to run my OpenClaw on a VPS. And the reason why you want to not have it on an actual computer is because if the computer turns off, then your OpenClaw is not working. Since it also has access to the computer itself, you know, it can get a bit dangerous if your security is compromised.
23:21So what I just recommend doing is hosting your own Openclaw on a VPS. Bluehost has a one click Openclaw setup now, which makes the whole Openclaw setup process really easy. It's built on their VPS infrastructure.
23:33As I'm making this video, I think if you select a two year plan, it gets as cheap as, like, $2 per month. So, yeah, from what I've seen, this is one of the most affordable VPSs that exist. And I'll put a link down below to that.
23:43Cool. So that's pretty much every AI model that you need to know about right now. Again, like I said, the AI space is moving so fast.
23:50So by the time you are watching this video, some of these model numbers might be a little bit different. But the most important thing, you guys, is that you actually use AI within your life. I highly highly recommend using AI in your personal life as well as your business life because it's gonna make you so much more efficient.
24:05If you're not taking advantage of AI, other people are, and . If you guys found this video helpful, make sure to like and subscribe. I make a ton of videos about AI, finance, business.
24:15My whole goal is to help you guys live a financially successful life. Thank you so much for your time, and I'll see you in the next video. Peace.
The Hook
The bait, then the rug-pull.
Charlie opens by name-dropping every famous AI brand in seven seconds - ChatGPT, Claude, Gemini, Grok, open source, video, images, agents - then lands the dream outcome ('work for you while you sleep') against a match-cut of someone asleep in bed. It's a promise-stack hook engineered to net every AI-curious viewer regardless of which brand they searched for.
Named ideas worth stealing.
00:50list
The Big Four chat models
- ChatGPT (best all-around)
- Claude (best writer + coder)
- Gemini (best value + multimodal)
- Grok (cheapest + real-time data)
Charlie's positioning of the four flagship chat models. Each gets one core sentence.
Steal forany 'choosing the right tool' explainer - the four-quadrant summary slide is screenshot-friendly and ends up shared on Twitter/IG
19:16list
Model-by-use-case routing for images
- ChatGPT image gen -> convenience + text-on-images
- Midjourney -> art / editorial polish
- Flux 2 -> photoreal / real skin and lighting
Three-tool routing matrix for image generation by output type.
Steal forany review-style longform where multiple competing tools each win at one specific job
15:00list
Open-source local-AI stack
- DeepSeek v3.2 (reasoning + coding, free web)
- Llama 4 Scout (10M context)
- Qwen 3.5 / Kimi K2 (open source benchmark leader)
- Mistral (runs on a single 24GB consumer GPU)
- Ollama (the runner that ties them together)
Charlie's recommended free-tier stack for people who don't want to pay for AI.
Steal forthe 'rent vs own your stack' framing for Mod Boss / $6 Stack content
How they asked for the click.
VERBAL ASK
23:50subscribe
“If you guys found this video helpful, make sure to like and subscribe. I make a ton of videos about AI, finance, business. My whole goal is to help you guys live a financially successful life.”
Soft, low-pressure. He earns it by ending on a universal motivational FOMO line ('other people are moving faster than you') before pivoting to the ask. Classic value-then-ask close.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
