Elon Musk’s Xai Launches Grok 4 Alongside A $300 Monthly Subscription

Elon Musk’s AI company, xAI, precocious connected Wednesday released its latest flagship AI model, Grok 4, and unveiled a caller $300-per-month AI subscription plan, SuperGrok Heavy.

Grok is xAI’s reply to models for illustration OpenAI’s ChatGPT and Google’s Gemini, and tin analyse images and respond to questions. In caller months, Grok has go much profoundly integrated into Musk’s societal network, X, which was precocious acquired by xAI — however, that’s besides put Grok’s misbehavior beforehand and halfway for millions of users.

The expectations are precocious for Grok 4. The latest AI exemplary from xAI will beryllium stacked up against OpenAI’s forthcoming AI model, GPT-5, which is expected to motorboat later this summer.

“With respect to world questions, Grok 4 is amended than PhD level successful each subject, nary exceptions,” said Elon Musk during a livestream Wednesday night. “At times, it whitethorn deficiency communal sense, and it has not yet invented caller technologies aliases discovered caller physics, but that is conscionable a matter of time.”

The motorboat of Grok 4 comes amid a tumultuous week for Elon Musk’s companies. Earlier connected Wednesday, Linda Yaccarino stepped down from her domiciled arsenic nan CEO of X aft astir 2 years pinch nan company. X has yet to denote her successor.

Yaccarino’s departure comes conscionable days aft Grok’s official, automated X relationship responded to users pinch antisemitic comments criticizing Hollywood’s “Jewish executives” and praising Hitler. xAI had to concisely limit Grok’s relationship and delete nan violative posts. In consequence to nan incident, xAI appeared to person removed a precocious added conception from Grok’s nationalist strategy prompt, a database of instructions for nan AI chatbot to follow, that told it not to awkward distant from making “politically incorrect” claims.

Musk and xAI’s leaders mostly avoided discussing nan incident, alternatively focusing connected Grok 4’s capacity and capabilities.

xAI launched 2 models connected Wednesday: Grok 4 and Grok 4 Heavy — nan second being nan company’s “multi-agent version” that offers accrued performance.

xAI claims that Grok 4 shows frontier level capacity connected respective benchmarks, including Humanity’s Last Exam— a challenging trial measuring AI’s expertise to reply thousands of crowdsourced questions connected subjects for illustration math, humanities, and earthy science. According to xAI, Grok 4 scored 25.4% connected Humanity’s Last Exam without “tools”, outperforming Google’s Gemini 2.5 Pro, which scored 21.6%, and OpenAI’s o3 (high), which scored 21%.

xAI claims that Grok 4 Heavy, pinch “tools,” was capable to execute a people of 44.4%, outperforming Gemini 2.5 Pro pinch tools, which scored 26.9%.

The nonprofit Arc Prize says that Grok achieves a caller state-of-the-art score connected its ARC-AGI-2 trial — different difficult benchmark that consists of puzzle-like problems wherever an AI has to place ocular patterns — scoring 16.2%. That’s astir doubly nan people of nan adjacent champion commercialized AI model, Claude Opus 4.

Alongside Grok 4 and Grok 4 Heavy, xAI launched its astir costly AI subscription scheme yet, a $300-per-month subscription called SuperGrok Heavy. Subscribers to nan scheme will get an early preview to Grok 4 Heavy, arsenic good arsenic early entree to caller features. The scheme is akin to ultra-premium tiers offered by OpenAI, Google, and Anthropic — however, xAI now offers nan astir costly subscription among awesome AI providers.

SuperGrok Heavy subscribers whitethorn get early entree to immoderate caller products xAI plans to motorboat successful nan coming months. The institution said Wednesday that an AI coding exemplary is coming successful August, a multi-modal supplier successful September, and a video procreation exemplary successful October.

The motorboat of Grok 4 comes aft a tumultuous week for Elon Musk’s societal media and AI companies.

Earlier connected Wednesday, Linda Yaccarino stepped down from her domiciled arsenic nan CEO of X aft astir 2 years pinch nan company. X has yet to denote her successor.

Yaccarino’s departure comes conscionable days aft an automated X relationship connected to Grok responded to users pinch antisemitic comments criticizing Hollywood’s “Jewish executives” and praising Hitler. On Tuesday, Elon Musk’s AI institution had to concisely limit Grok’s relationship from responding to users, and deleted nan violative posts.

In consequence to nan incident, xAI appeared to person removed a precocious added conception from Grok’s nationalist strategy prompt, a database of instructions for nan AI chatbot to follow, that told it not to awkward distant from making “politically incorrect” claims.

Musk and xAI’s leaders did not make overmuch of immoderate mention of nan incident, alternatively focusing connected Grok 4’s capacity and capabilities.

Despite Grok’s frontier-level capacity connected benchmarks, it’s difficult for xAI to move past these mishaps arsenic it tries to transportation Grok to businesses arsenic a existent contender to ChatGPT, Claude, and Gemini.

xAI is releasing Grok 4 done its API successful an effort to get developers to build applications pinch nan model. The institution notes that xAI’s endeavor assemblage is only 2 months old, however, it plans to activity pinch hyperscalers to make Grok disposable done their unreality platforms.

Whether businesses are fresh to adopt Grok, flaws and all, remains to beryllium seen.