Elon Musk’s next AI prototype, XAI, may be about to be released.
During the weekend, numerous Users On x, incl Reverse engineer Alexei Shabanovaccess to Grok 3, the successor to Xai’s current model, Grok 2, was reported to be accessed via .
According to users, Grok 3 manages to be a success answer Puzzles and Field requests Such as “Create HTML and JavaScript code for a roulette wheel casino.” The model wasn’t perfect. In the code for the roulette wheel, Grok 3 missed some details – and made a programming error.
BREAKING 🚨 Grok 3 understands the subtle puzzle that all the open AI models have failed to get right. The only other model to get this question right is the Deepseek R1 🤯
Grok 3 is the O1 hit? He has xai Caught this quickly? pic.twitter.com/wlspm00qhk
– Chris (@chatgpt21) January 26, 2025
Shabanov said he got the Grok 3 to detect its system prompt, which is the set of instructions that determine how the model behaves. One interesting detail is that the claim appears to explicitly state Donald Trump as the 47th President of the United States, which could be an extreme solution to Grok’s political hallucinations.
Grok 3, which has been in development for several months, was optimistically scheduled for launch last year, according to Musk — but it missed that deadline. Early this month, musk He said That Grok 3 has completed pre-training, a critical stage in the model development cycle, and that expected Grok 3 to arrive in January or early February.
XAI is using its massive data center in Memphis — a data center containing about 100,000 GPUs — to train Grok 3. In a mail In X, Musk claimed that the Grok 3 was trained with “10x” more compute than the Grok 2.
Grok, Xai’s answer to models like Openai’s GPT-4O and Google’s Gemini, can analyze images, respond to questions, and air out a number of features on X. Grok recently gained standalone apps, and may soon get a “”Sound modeThis reads the form’s responses aloud.
When Musk announced Grok about two years ago, he developed the chatbot as edgy, unfiltered, and anti-“woke”—generally willing to answer controversial questions that other AI systems wouldn’t. He delivered on some of that promise. For example, Grok is asked to be vulgar, you’ll happily oblige, and spews profanity and colorful language that you won’t hear from chatgpt.
But Grok as it exists today Hedging On political issues, they will not express themselves certain limits. In reality, One study It found Grok to lean to the political left on topics like transgender rights, diversity programs, and inequality.
Musk blamed the behavior on Grok’s training data – public web pages – and pledge To “turn Grok closer to political neutrality”. Evidence of Grok’s “undefined mode” has recently surfaced, which will provide responses “intended to be objectionable, inappropriate, and offensive,” according to an FAQ page on Xai’s website.
Musk also said that Grok 3 training data includes Files from court casesImprove their ability to understand legal topics.
