Introducing Grok by Elon Musk – Explore Its Capabilities and Find Out the Price Tag

Key Takeaways

• Elon Musk’s AI startup xAI has unveiled Grok, a large language model aiming to compete with other leading models like OpenAI’s GPT and Anthropic’s Claude.
• Grok is designed to answer questions with wit and sarcasm, showcasing Musk’s influence on its tone and personality.
• While still in beta testing, early results suggest that Grok outperforms other models in machine learning benchmarking, and it came second in a test against OpenAI’s GPT-4.

The battle for AI supremacy rages on, with Elon Musk unveiling Grok, the first large language model from his fledging AI startup, xAI.

Grok becomes a new addition to a busy generative AI world, aiming to compete with market-leading opposition OpenAI and its GPT models, along with Anthropic’s Claude, Meta’s Llama 2, Google’s PaLM 2, and countless other LLMs.

Musk has talked xAI’s Grok up, but does it have what it takes to compete?

What Is xAI’s Grok?

Despite Musk regularly reminding the world that AI could be bad, that it could create a world without jobs for regular folks (if you believe Musk’s take on this is benign, I have a bridge to sell you), and that the entire focus of AI needs greater governance, Musk proudly proclaimed on X that “In some important aspects, it [Grok] is the best that current exists.”

Grok-1, then, is a large language model similar to OpenAI’s GPT-3.5 and GPT-4. xAI’s generative AI chatbot also currently appears to be called Grok (similar to how Anthropic’s Claude model is also the same as its AI tool).

But while Grok will answer questions in a conversational manner, Musk appears to have had some influence on Grok’s tone, withxAI’s official Grok announcement stating:

Grok is designed to answer questions with a bit of wit and has a rebellious streak, so please don’t use it if you hate humor!

Or, as Musk himself said, Grok “is based & loves sarcasm. I have no idea who could have guided it this way.”

For any casual observer, Musk’s chortling about being based is anything but, landing more on the toe-curling plane of discomfort, delivering a visceral sense of embarrassment. But for Musk’s legion of fans, xAI’s Grok is the long-awaited “answer to the monstrosity of WokeGPT,” a reference to the boundaries and perceived neutering of ChatGPT’s capabilities by OpenAI.

How Does Grok Perform? Is Grok Better Than ChatGPT?

Given only a minute number of users have been given access to Grok so far, most of what we know about its performance comes from xAI. The company was keen to stress that “Grok is still a very early beta product” and that its current performance is “the best we could do with 2 months of training.”

Even so, Grok appears to be on the right track. According to xAI’s research, its prototype Grok-0 model was trained using 33 billion parameters and outperformed the Llama 2 70 billion parameter model in standardized machine learning benchmarking and testing. After some tweaks (and presumably an expanded set of parameters), Grok-1 outperformed OpenAI’s GPT-3.5 (around 175 billion parameters).

However, in a further boon to Musk and Grok, during a final round of testing on the 2023 Hungarian National High School Final Exams in Mathematics (yes, obscure, but purposefully so to avoid the chance the data has been uploaded to an existing LLM, which would give it an advantage), Grok came second only to OpenAI’s GPT-4.

Musk has also observed that Grok’s direct internet access will give it an advantage over other generative AI tools that lack access.

Currently, it’s difficult to say whether Grok is better than ChatGPT, and it’s clear that Grok is speeding forward based on the work already laid out by other AI research companies.

How Much Will Grok Cost? Who Can Use Grok?

Grok is still currently in early testing. Limited beta access has been granted to a small number of X Premium Plus users, which currently costs $16 per month. Once Grok leaves its beta period, it will become available to all X Premium Plus subscribers, dangling a weighty AI-shaped carrot over X verification.

It’s currently unknown if Grok will come with a free access tier, which all other major generative AI tools offer. For example, OpenAI keeps its less-powerful ChatGPT-3.5 model free, while Anthropic offers a free version of Claude. Restricting Grok to X’s verified paying customers is unlikely to pay the bill, especially given X’s drop off in user numbers since Musk’s takeover.

I would expect xAI to release an earlier version of Grok for free, or at least, much cheaper, in the hope that it encourages users to upgrade to the full version. Another issue that may stand in the way of Grok uptake is its current direct link to X Premium Plus. There may be users that want to subscribe to Grok but don’t want X verification or one of Musk’s paid-for Blue Ticks next to their name—something else for Musk, X, and xAI to wrestle with.

Musk has indicated that a separate version of Grok may launch in the future, but there is little information on that at the current time.

What’s Next for xAI’s Grok?

xAI’s development of Grok ticks off another goal for Musk, who was once part of the OpenAI team. He’s long rallied against OpenAI’s perceived censorship of ChatGPT, arguing that limiting its capabilities is limiting what people can and should be able to do with such powerful AI tools.

What comes next for Grok is still unknown. While it’s clear some limitations are built into Grok, the real test for its boundaries will only become clear once the public gets its hands on it.