Elon Musk unveils Grok 3 chatbot
Add Axios as your preferred source to
see more of our stories on Google.
Screenshot: xAI
Elon Musk on Monday announced Grok 3, the next version of xAI's chatbot, touting it as "an order of magnitude more capable than Grok 2."
Why it matters: The move comes as Musk has been engaged in a bitter rivalry with OpenAI that has included lawsuits, a war of words and — most recently — an unsolicited $97.4 billion takeover bid.
The big picture: During a video livestream, Musk and his team said the goal of xAI and Grok is to understand the nature of the universe.
- The video discussion started with benchmarks, details about xAI's new data center and the desire to create a chatbot that didn't need to be politically correct.
- Musk and xAI said that Grok 3 contains advanced reasoning abilities and will continue being improved through reinforcement learning.
- "We're seeing the beginnings of creativity," Musk said.
In a series of demos, Musk and his colleagues asked Grok 3 to solve a physics problem and create a game that combines Bejeweled and Tetris.
- "I've played the game," Musk said, of an earlier creation using that prompt. "It's pretty good."
Between the lines: The Grok 3 API, which includes the reasoning model and DeepSearch, will launch "in the coming weeks."
- The company described DeepSearch as both a new kind of search engine and an early version of agent-like capabilities.
- A new "SuperGrok" subscription will provide advanced access to Grok's features.
- There is also a mini version of the reasoning model, the company said.
Yes, but: xAI said Grok 3's reasoning model is in beta and is still being trained.
- "This is kind of a beta where you should expect some imperfections at first," Musk said. "If you want a more polished version, perhaps wait a week."
- A planned voice mode was pulled from Monday's launch, Musk confirmed in a post on X.
- "Voice mode is still a little patchy, so probably launches in about a week, but it's awesome," Musk wrote.
