Grok 3 unveiling Grok 3 unveiling
Elon Musk joined xAI engineers and experts Monday night to unveil the much-anticipated Grok 3 AI chatbot. xAI on X / Live Broadcast Screenshot

KEY POINTS

  • Lex Fridman was mind-blown by Grok 3 after 'extensively' testing the chatbot's early version
  • Robert Scoble hailed the AI assistant's learning speed, which he noted was faster than others
  • Brian Roemmele compared Grok 3's impressive reasoning capabilities to ChatGPT's 'lame' outputs
  • Grok 3 is now No. 1 across all categories on AI chatbot benchmarking platform LM Arena

Elon Musk's artificial intelligence startup xAI has unveiled its latest flagship model, Grok 3, and so far, industry experts are impressed with the AI assistant's capabilities, with some praising the chatbot's "reasoning" prowess.

The tech titan previously said Grok 3 is the "smartest AI on Earth," having been trained on synthetic data and was built with a self-correction feature to avoid making errors.

Industry experts who were given early access to the latest iteration of the Grok AI chatbot and those who kept up with the live demo on X have shared their thoughts on the model. So far, things are going all too well for xAI.

Tech Experts Impressed with Grok 3's Learning Benchmarks

Prominent American computer scientist Lex Fridman revealed he was able to use the chatbot "extensively" and ahead of the model's unveiling.

"My mind is blown, very impressive model," he said, sending his congratulations to Musk and the xAI team for bringing the AI assistant to life.

Well-known AI pioneer Robert Scoble was also impressed. "The thing to really pay attention to in AI is learning speed. And @xai is learning way faster than any other," he wrote, along with a screenshot of Grok 3's learning benchmarks compared to the earlier Grok iterations, Google's Gemini 2 Pro model, Chinese DeepSeek, Claude 3.5 Sonnet, and OpenAI's ChatGPT4.

Software expert and tech executive Aaron Levie said Grok 3 was coming out "very strong." He said the model's learning abilities are "great proof that the scaling laws are not, in fact, over."

DeepSeek's V3 scored 40 on the Coding benchmark, ChatGPT scored 43, while Grok 3 logged a score of 57. The numbers are even higher on Science (GPQA), with Grok 3 scoring 75, DeepSeek scoring 65, and ChatGPT settling at 50.

Grok 3 Ranks 1st on LM Arena

Within hours since the early version of Grok 3 went out on popular AI chatbot comparison platform LM Arena, Musk's AI assistant took the top spot, besting Google's Gemini 2.0 Flash, ChatGPT, and several other chatbots.

According to LM Arena, Grok 3 was the first model to break a score of 1,400 and it is also the top chatbot "across all categories, a milestone that keeps getting harder to achieve."

Grok 3 Has Better Reasoning Skills Than ChatGPT?

The xAI chatbot's LM Arena rankings on various benchmarks are already impressive as is, but for some industry experts, Grok 3's reasoning abilities are what make it stand out.

"Grok 3 has one of the best reasoning engines available today," said tech expert and researcher Brian Roemmele. He said the model shows "significantly more than 'OpenAI' lame PowerPoint 'reasoning' output."

Scale AI CEO Alexander Wang had similar thoughts, noting how Grok 3 scored "impressively on pretraining and reasoning" evaluations.

For Amjad Masad, the CEO of AI agent developer Replit, xAI's accomplishment with Grok 3 shouldn't be ignored, considering "how late in the game they started."

The model's unveiling came after Chinese startup DeepSeek's AI assistant rocked the tech world due to its model that it supposedly developed on a small budget.

Originally published on IBTimes