00:00:02 AI is everywhere these days.

00:00:05 Whether it's chat bots, search engines, or content creation tools, AI models are helping us deliver fast, and most of the times,

00:00:13 accurate and meaningful responses.

00:00:16 But with so many AI models out there, picking the right one can feel like choosing a movie or series on Netflix.

00:00:23 Way too many options, and all of them seem good until they aren't.

00:00:27 And while you already may know some, let's take a look at the most famous AI models, one of which will choose to be our own AI generation model to generate

00:00:37 AI answers within Demflow.

00:00:39 First one on the list, OpenAI.

00:00:42 While its name is OpenAI, it is not open.

00:00:45 Jokes aside, it's one of the most well-known language models out there, capable of understanding and generating human-like text with high accuracy.

00:00:54 It's great for chatbots, content creation, and answering complex coding questions.

00:01:00 Then there is Meta's Llama, an open-source LLM designed for efficiency and scalability.

00:01:07 It's often used in research and enterprise applications, often requiring private AI solutions.

00:01:14 Google, of course, also has their own LLM called Gemini, Previously known as BARD, Gemini is a Google's answers to AI chatbots.

00:01:23 It's deeply integrated with all Google services, like YouTube, for example, making it useful for real-time data and research.

00:01:31 And there are two new major AI LLMs.

00:01:34 one of which is called Mistral, similar to other models.

00:01:38 It's also lightweight in performance, but still delivers quality responses.

00:01:42 And finally, the last one in the list, but maybe the biggest of them all, often called the destroyer of Nvidia stocks and the open AI killer,

00:01:51 the name is DeepSeek.

00:01:53 developed by a few developers from China as a personal project.

00:01:57 Still, it's one of the top emerging LLMs designed both for general purpose language understanding and specialized code generations.

00:02:06 It performs well in logical reasoning, text comprehension, and programming tasks.

00:02:10 And it's been gaining a lot of open source attention recently.

00:02:14 How would it not?

00:02:15 It managed to do what OpenAI did with billions of dollars.

00:02:18 And while these are the best models right now, By the time you're watching this video, a new one might appear.

00:02:24 There's a crazy amount of AI hype happening, and you should never integrate an AI model in your application just for the sake of integrating it.

00:02:31 You want to figure out if it actually makes sense to do so.

00:02:34 Once you decide that to be the case, you have to choose which one you want to use.

00:02:38 And sometimes your app might require you to switch between different AI models based on the use case or the features you want to give to your users.

00:02:46 Of course, you don't want to create separate integrations for each one of these models.

00:02:51 as that would result in redundant code, high maintenance costs, and too many conditional statements switching between different models.

00:02:59 So there has to be a better way, right?

00:03:01 A way that would allow you to dynamically switch AI models with just a single line of code.

00:03:06 Well, thankfully there is one, and it integrates with Next.js perfectly.

00:03:11 It's the Vercelles AI SDK.

00:03:14 It provides unified APIs for integrating different AI models.

00:03:18 So instead of writing separate implementations for each AI, like OpenAI, Entropic, HuggingFace, and so on, we can use a single function call and change

00:03:27 the model by modifying one of the parameters.

00:03:30 Pretty cool, right?

00:03:31 That's exactly what we'll do today.

00:03:33 I'll teach you how to use Vercell's AI SDK to implement the main logic.

00:03:37 And then you'll be free to choose any AI model of your choice while keeping the code base super clean.

00:03:43 So let's do that in the next lesson.