Elon Musk’s xAI launches Grok 4 alongside a $300 monthly subscription

The expectations are excessive for Grok 4. The newest AI mannequin from xAI can be stacked up in opposition to OpenAI’s forthcoming AI mannequin, GPT-5, which is predicted to launch later this summer time.

“With respect to educational questions, Grok 4 is healthier than PhD degree in each topic, no exceptions,” stated Elon Musk throughout a livestream Wednesday night time. “At occasions, it could lack widespread sense, and it has not but invented new applied sciences or found new physics, however that’s only a matter of time.”

The launch of Grok 4 comes amid a tumultuous week for Elon Musk’s corporations. Earlier on Wednesday, Linda Yaccarino stepped down from her function because the CEO of X after roughly two years with the corporate. X has but to announce her successor.

Yaccarino’s departure comes simply days after Grok’s official, automated X account responded to customers with antisemitic comments criticizing Hollywood’s “Jewish executives” and praising Hitler. xAI needed to briefly restrict Grok’s account and delete the offensive posts. In response to the incident, xAI appeared to have removed a recently added section from Grok’s public system prompt, a listing of directions for the AI chatbot to observe, that instructed it to not shrink back from making “politically incorrect” claims.

Musk and xAI’s leaders largely averted discussing the incident, as a substitute specializing in Grok 4’s efficiency and capabilities.

xAI launched two fashions on Wednesday: Grok 4 and Grok 4 Heavy — the latter being the corporate’s “multi-agent model” that provides elevated efficiency. Musk claimed that Grok 4 Heavy spawns a number of brokers to work on an issue concurrently, after which all of them examine their work “like a examine group” to search out one of the best reply.

xAI claims that Grok 4 exhibits frontier degree efficiency on a number of benchmarks, together with Humanity’s Final Examination— a difficult check measuring AI’s capability to reply 1000’s of crowdsourced questions on topics like math, humanities, and pure science. Based on xAI, Grok 4 scored 25.4% on Humanity’s Final Examination with out “instruments,” outperforming Google’s Gemini 2.5 Professional, which scored 21.6%, and OpenAI’s o3 (excessive), which scored 21%.

xAI claims that Grok 4 Heavy, with “instruments,” was in a position to obtain a rating of 44.4%, outperforming Gemini 2.5 Professional with instruments, which scored 26.9%.

The nonprofit Arc Prize says that Grok achieves a new state-of-the-art score on its ARC-AGI-2 check — one other troublesome benchmark that consists of puzzle-like issues the place an AI has to establish visible patterns — scoring 16.2%. That’s almost twice the rating of the following finest industrial AI mannequin, Claude Opus 4.

Alongside Grok 4 and Grok 4 Heavy, xAI launched its most costly AI subscription plan but, a $300-per-month subscription known as SuperGrok Heavy. Subscribers to the plan will get an early preview to Grok 4 Heavy, in addition to early entry to new options. The plan is just like ultra-premium tiers provided by OpenAI, Google, and Anthropic, however xAI now presents the most costly subscription amongst main AI suppliers.

SuperGrok Heavy subscribers could get early entry to some new merchandise xAI plans to launch within the coming months. The corporate stated Wednesday that an AI coding mannequin is coming in August, a multi-modal agent in September, and a video era mannequin in October.

xAI is releasing Grok 4 by means of its API in an effort to get builders to construct purposes with the mannequin. The corporate notes that xAI’s enterprise sector is simply two months outdated, nonetheless, it plans to work with hyperscalers to make Grok accessible by means of their cloud platforms.

Regardless of Grok’s frontier-level efficiency on benchmarks, it could show troublesome for xAI to maneuver previous its latest mishaps because it tries to pitch Grok to companies as an actual contender to ChatGPT, Claude, and Gemini. Whether or not companies are able to undertake Grok, flaws and all, stays to be seen.