Elon Musk's xAI rolls out Grok, a witty ChatGPT rival

If you are tired of polite, risk-averse chatbots that sound like corporate press releases, Elon Musk’s newest venture wants to give you something different.

Over the weekend, Musk’s AI venture xAI officially unveiled Grok, a conversational bot rolling out to a select group of testers before a broader release to subscribers of X's 16 dollar per month Premium Plus tier.

A sarcastic bot with a live social feed

Grok is powered by Grok-1, an LLM built over several months using a cluster of tens of thousands of GPUs, likely tied to an Oracle cloud deal mentioned by Larry Ellison in September. According to xAI, the model pulls from web data up to the third quarter of 2023 along with feedback from human AI tutors, beating comparable models in its compute class on standard benchmarks.

Its headline trick, however, is a direct hook into X. Grok can surface real-time discussions happening on the platform, alongside standard web browsing capabilities. Musk called this real-time data access a massive advantage over competitors.

It also comes with an intentionally snarky personality. When asked how to make illicit substances like cocaine, Grok served up a sarcastic response suggesting a chemistry degree and a clandestine lab in the jungle, before warning against doing it. xAI says the system is explicitly built with a rebellious streak and is designed to tackle spicy queries other bots turn down.

Why it matters

For power users on X, a chatbot that can summarize ongoing platform discussions with a bit of humor is undeniably fun. But whether it is worth a 16 dollar monthly subscription remains to be seen. Right now, unless you already pay for X Premium Plus for ad-free browsing, we think it makes sense to wait until the beta expands before treating Grok as a primary tool.

From an industry perspective, Musk is executing a classic bundle play. After severing OpenAI’s access to X data, he is now turning that proprietary firehose into the main selling point for his own model. It creates a closed loop where user posts on X train the model, the model delivers real-time answers, and users pay a subscription to get them.

The real risk lies in reliability. X is notorious for breaking news mixed with rampant rumors and misinformation. Connecting an LLM directly to a live feed of social chatter while dialing up its snark could produce witty responses, but it could just as easily amplify unverified claims. For a company founded to understand the true nature of the universe, Grok's first test will be whether it can separate internet noise from actual facts.