xAI launches Grok 3 for X's Premium+ subscribers and introduces new SuperGrok subscription
Elon Musk's xAI has unveiled Grok 3, branded as "the smartest AI on Earth," with a live demo highlighting its capabilities. Grok 3 competes with or surpasses OpenAI’s GPT-4o and Google’s Gemini 2.0 Flash Thinking in various AI performance benchmarks. Trained on over 100,000 Nvidia H100 GPUs, it accumulated 200 million GPU hours, tenfold the computing power used for Grok 2.
Grok 3 features an advanced reasoning model for more accurate responses and a "big brain" mode for complex queries, advantageous for math, science, and coding tasks. DeepSearch integration enables real-time web data analysis, including internet and X scanning. A Grok 3 Mini version offers faster response times.
Accessible via a standalone app, some internal processes are obscured to prevent AI distillation. Initially available to X’s Premium+ subscribers, additional AI features will require a SuperGrok subscription. With Grok 3's launch, X has also increased its Premium+ pricing by nearly 50% from its original value, reaching $40/month or $395 annually, up from $22/month. Premium+ is X’s highest-priced subscription, providing almost ad-free browsing (with some exceptions). X’s subscription tiers now cost $3/month for Basic, $8/month for Premium, and $40/month for Premium+, the only tier currently granting access to Grok 3 for now.



Comments
They should finally release an App for it
There are already standalone apps for iOS and Android. You can find them on the app page here:
Grok
Elon for the win again. Haters gonna hate.
Okay person with N*zi dogwhistle in their name, whatever you say! Hope his boot is delicious :P
Right? just like those Path of Exile 2 wins he was scoring! But don't let online dickriding get in the way of your prayers to him. Dork Maga will know.
Competes with an one-year old OpenAI model but twice more expensive? It has to be trained with US Treasury data to worth anything.
Not true. Grok 3 outperforms GPT-4o by a wide margin: it's about 30% faster, 25% cheaper, more updated, self-corrects with roughly 30% fewer hallucinations than GPT-4, and has reasoning capabilities that match or exceed those of OpenAI's O3-pro model ($200/month).
Your claims are wrong or partially untrue: https://www.perplexity.ai/search/critically-analyze-and-fact-ch-z0pMab4SSWe5rcT8lW.hjA
The numbers I gave you were my own verified calculations, not an AI.