In a week where Elon Musk seems determined to keep tech journalists typing, xAI has unveiled Grok 3: its latest AI model that’s been training on what might be the world’s most powerful computer cluster.
The launch comes as Musk juggles a controversial bid for OpenAI and faces scrutiny over his ‘DOGE’ government efficiency department, proving that even tech billionaires sometimes struggle with work-life balance.
A Colossal Achievement
At the heart of Grok 3’s development is “Colossus,” and the name isn’t just Silicon Valley showmanship.
The massive cluster of 200,000 GPUs took 214 days to reach full capacity, with xAI’s developers claiming that building the infrastructure proved more challenging than developing the AI model itself!
The system came together in two distinct phases: 122 days of synchronous training on 100,000 GPUs, followed by 92 days of scaling up to the full 200,000. That’s a lot of computing power, and it shows in the results. Grok 3’s base model has raised eyebrows with its performance across standard AI benchmarks, particularly in mathematics (AIME), science (GPOA), and coding (LCB) tests. A specialized “Reasoning Beta” variant pushed even further, achieving 93% on the AIME 2025 benchmark, outperforming competitors that scored below 87%.
Take a look at Grok 3 compared to its closest competitors:
Image from Grok 3 launch
Impressive Capability, but Too Late to the Table?
The model, previously known by the codename “Chocolate,” has proved particularly sweet in blind tests, achieving the highest ELO score in the LLM Arena. This means users preferred its answers over other AI models without knowing which was which; perhaps the most honest way to measure AI performance short of a Turing test over dinner.
Industry insiders were impressed:
I got to use Grok 3 extensively (early). My mind is blown, very impressive model Congrats to Elon and the team for bringing it to life
— Lex Fridman (@lexfridman) February 18, 2025
However, while xAI’s demonstrations showcased impressive capabilities, many of them had a distinct déjà vu quality. Physics problem-solving, game code generation, and web research tools—all impressive, but this is all territory previously charted by ChatGPT, Claude, and Google’s Gemini.
It’s less about breaking new ground and more about catching up to the competition, albeit with style. We are creatures of habit, so whether Grok can convince dedicated, paid-up users of other genAI engines to make the switch remains to be seen.
Features and Future Plans
The company has announced a suite of upcoming features that suggest xAI isn’t content with just matching its rivals. DeepSearch promises to revolutionise web research, while new AI voice interactions aim to make conversation with Grok 3 feel more natural than ever. Developer API access and audio transcription capabilities are on the horizon, and there’s even an AI gaming studio platform in the works.
If you’re excited about trying Grok 3, prepare your wallet. X has marked the spot for Premium+ subscribers with a chunky price hike—nearly $50 monthly or $350 annually; more than doubling the previous rate.
Some of Grok 3’s advanced features, including “deep search” and “reasoning,” will require a separate ‘SuperGrok’ plan, available through the Grok app.
The Bigger Picture
The timing of Grok 3’s release adds another layer to an already complex tech narrative. While xAI was putting the finishing touches on its AI model, Musk was making waves with a $97.4 billion conditional bid for OpenAI—a move that prompted OpenAI’s CEO Sam Altman to counter with a cheeky offer to “buy twitter for $9.74 billion.” Meanwhile, Musk’s Department of Government Efficiency (DOGE) faces legal challenges from 14 state attorneys general and a Treasury Department audit, suggesting that efficiency might be in the eye of the beholder.
Looking ahead, xAI has announced plans to expand its computing infrastructure fivefold, potentially creating what would be the world’s most powerful GPU cluster. It’s an ambitious goal that shows xAI isn’t just here to take part, it’s here to (try to) take over.
Whether Grok 3 catches on commercially like its more well-established counterparts – especially with such a decisive figure as Elon Musk at the helm – remains to be seen, but one thing is clear: xAI is determined to be a frontrunner in the grueling race for AI supremacy.
With its massive computational resources and aggressive development timeline, Grok 3 might just help xAI grok its way to the top of the AI hierarchy—assuming, of course, that Musk can keep all his other plates spinning in the process.
For now, we’ll have to wait and see if this latest offering is truly groundbreaking, or just another step in the ongoing AI evolution that’s moving at breakneck speed.
Want to get involved in the discussion? Join our Linkedin group here and have your say.
