What is Grok AI? xAI's model vs ChatGPT and Claude

Grok — watch on YouTube
5 min

In 30 seconds

  • Grok's real advantage is live information: it runs on X, which puts it ahead of the other models on data that is minutes old.
  • xAI has shipped about five generations of Grok in three years, but the model has plateaued since Grok 4 Heavy in July 2025.
  • Grok is one of the first models that can take a video as input instead of chopping it into frames and reading the frames as images.

Grok is the AI model built by xAI, Elon Musk’s AI company, and it runs on X. That is its edge over ChatGPT, Claude and Gemini: real-time information, data that is minutes old. On most other measures it has been standing still for about a year.

Who owns Grok?

Elon Musk, through xAI. He co-founded OpenAI alongside Sam Altman and the rest of the team, left the board in 2018, and in early 2023 started xAI as a competitor. Grok came out of that. The stated mission is to build the most truth-seeking AI.

Three events, in order. xAI starts in 2023. Two years later it acquires X. A year after that, xAI and X together are acquired by SpaceX, which is also Musk’s. So Grok belongs to xAI, and xAI belongs to a rocket company.

Every Grok model, from Grok 1 to Grok 4 Heavy

The timeline is compressed next to everyone else’s. OpenAI has been shipping since 2018. Grok started at the end of 2023 and has produced about five generations in three years.

When Model What it added
End of 2023 Grok 1 Launched to paying X subscribers as a pro feature, trained on X discussions and the news breaking there
About a year later Grok 2 Started competing with GPT-4o on reasoning and image generation
Not dated in the video Grok 3 Reasoning as the main pitch
Not dated in the video Grok 4 Passed Humanity’s Last Exam and other top benchmarks
July 2025 Grok 4 Heavy The last real jump. Flat since
Expected end of year Grok 5 Back in a race with Claude and GPT

When is Grok 5 coming out?

By the end of the year, going on what the market expects rather than an announced date. Until it lands, the top of the line is still Grok 4 Heavy from July 2025, and nothing much has moved in the year since. Grok 5 is the release that decides whether xAI is racing Claude and GPT again or watching them.

What Grok is best at

  1. Real-time information. Grok has proprietary access to what is being posted on X, which puts it ahead of the other models on live data. Finance and trading products get built on top of it for that reason.
  2. Price. xAI prices aggressively. Grok is cheaper than the rest.
  3. Context window. Grok arrived with a 1 million token context window, which was a lead until the other models matched it. A 2 million token version exists, the largest context window you can use right now.
  4. Video input. Grok was among the earliest models able to take a video in. Models that generate video usually cannot read one: they chunk the file into frames and analyse the frames as images. Grok reads the video.
  5. Unfiltered output. Grok’s answers come back less filtered than the alternatives, so when you want an objective read on something rather than a careful one, Grok is the one to ask.

Grok vs ChatGPT, Claude and Gemini

Grok ChatGPT Claude Gemini
Real-time data Live X feed Behind Grok Behind Grok Behind Grok
Price Cheapest of the frontier models Dearer Dearer Dearer
Context window 1 million, 2 million available Caught up to 1 million Caught up to 1 million Caught up to 1 million
Video input Reads the video Frames only Frames only Frames only
Filtering Least filtered More filtered More filtered More filtered

The video treats the other three as one group on these axes, so the last three columns read the same by design. Four wins for Grok, and one loss underneath them all: general capability, flat since July 2025 while the others kept shipping. Grok is a specialist you reach for on a specific job. Moving the rest of your work onto it costs you more than the price saves.

Is Grok free?

Not on the evidence of how it shipped. Grok 1 went out as a pro feature to paying X subscribers, so the way in was an X subscription rather than a free tier. xAI prices below the other frontier models, which is one of the reasons people build on it. The video gives no plans and no per-token figures, so take it as cheaper, not as free.

Two routes in. The subscriber product inside X, which is the one an operator uses. And building directly on the model, which is what the finance and trading products do when they need data measured in minutes.

Where xAI is heading

xAI’s future is the infrastructure layer. It has signed deals with Anthropic to lease out its data centres, which means xAI hardware running Claude. Grok itself probably gets retired within a couple of generations: 5, 6, perhaps 7, and then that is the end of the line.

Musk’s ambition is to put data centres into orbit and run the inference up there, rather than in a desert where the chips have to be cooled. Among AI companies that is an unusual place to point a roadmap, and it tells you where the attention inside xAI is going.

Should you use Grok

For the everyday work in a solar or HVAC business, writing quotes, summarising site notes, handling after-hours enquiries, Grok gives you nothing the model you already pay for does not.

Reach for it when the job needs one of its specifics: live data measured in minutes, raw video handed to a model instead of screenshots pulled out of it, an objective answer instead of a careful one, or a price per call low enough to change what you can afford to run at volume. A property manager pushing a walkthrough video of a vacated unit straight into a model, with no frame extraction step in between, is using the one capability Grok has that the others mostly do not.

Keep whatever you build on it portable. A model that has plateaued for a year, from a company whose owner is planning data centres in orbit, is not the thing to hard-wire into your dispatch process.

Full transcript

Expand

Hey, my name is Mahan. I'm the CMO of Nairon. My name is Luka and I'm the CEO of Nairon. Welcome to our AI model series where we cover some of the most popular AI models out on the market right now. In today's episode, the golden word rock. So, Luka, take it from there.

All right, so a quick recap. In the previous video on OpenAI, we discussed how Elon Musk essentially co-founded OpenAI alongside Sam and the rest of the team and then left the board in 2018. in early 2023 start a competitor of OpenAI called XAI and that's where Grock initially came out of. What's important to note is three main events. First of all, the start of XAI starts in 2023. 2 years later it acquires Twitter X in this case and then only a year later the entire company XAI and Twitter gets acquired by SpaceX which is Elon Musk's company. And all this under the one mission of creating the most truth seeking AI. And so what we'll see here with the timeline, it's very compressed.

If we look at OpenAI's timeline spanning all the way from 2018 till today, we look at Grock. Grock only started at the end of 2023 and 3 years later, we have about five different models, five different generations of models that have come out. So if we take a look at the first one, um, Grock 1 actually launches to paying subscribers of X sort of as a pro feature. Um, mainly trained on a lot of the X discussions and and news that it's coming out on X. Within a year or so, Gro 2 comes out and now Gro 2 actually starts competing with the likes of GPT 4.0 and the other reasoning models because it has a lot of good strong reasoning models and the image generation was quite quite good. And that's where Grock 3 really starts competing with the other models mainly because of its reasoning functionality.

Grock 4 is where um it starts passing some of the biggest benchmarks like the humanities last exam for example and in July 2025 that's where they launched Grock 4 heavy and sort of have seen a plateau since then. And so what we really see is over the past one year or so, there's not been a lot of movement with Grock and we're anticipating by the end of the year, Grock 5 comes out um and could start competing back again with the likes of Claude and GPT.

Okay. And it's obviously one of the younger models out there, especially compared to what's we consider Frontier at the moment. So let's dive deep into exactly what Grock really shines at compared to everything else. A lot of the use cases with Grock specifically is its real time information. So because it's built on X and has a lot of proprietary data on what's coming out through X um it's actually ahead of the curve in terms of realtime data compared to the other models. Um and so a lot of use cases for example in finance and trading a lot of these products are built on top of Grock because it's so frontier in terms of the time to use really. Um another one is it's it's quite aggressive on pricing.

So it's cheaper than the rest of the models. Um it has a very long context window. It kind of started out with the 1 million context window but then sort of all the models have already kind of come up to there. Um there is a 2 million parameter model as well which you can use as well um which is probably the largest right now. It's one of the first models to actually intake video.

So other models that can produce video cannot intake video. The way that the other models would need to do is they would need to chunk up the videos in frames and analyze the video in terms of images. This one can intake videos and then in terms of its output, it's a lot less filtered. So if you really need an objective perspective on any given answer, um it's probably the best model to do that.

And so what do you see as the the future of XAI and Grock in the next coming years? So the future of XAI I believe is more on the infrastructure layer. um they actually have signed deals with Anthropic to lease out their data centers to provide the power to run um actually Claude and Claude's models. I think Grock probably gets retired in a couple of generations. They'll probably ship out five and six, maybe even seven and sort of keep it at that. Um and they'll probably go more into the infrastructure layer. I I think Elon Musk's vision is to actually send these data centers into the orbit and run a lot of the infrastructure and inference in the actual orbit itself, not on Earth, not in like the desert and have to like cool the chips and all of that.

I think the whole mission of XAI is to send these data centers into the orbit where then they could run there at, you know, whatever temperatures they need to do. Interesting. And that's very unique when it comes to AI companies. Um, but that being said, guys, I hope you enjoyed the video. In the meantime, you can subscribe to our newsletter. Links down in the description below. You can also connect with us on LinkedIn. We're very active on a daily basis. And until next time, we'll see you in the next video.

Next in AI Models ExplainedLlama
Ready to put AI employees to work?
Book a call
256-bit SSL Secured
© 2026 Nairon, Inc. All rights reserved.
PrivacyTerms & ConditionsCookie PolicyAcceptable Use