Elon Musk’s xAI has officially released Grok 4, the most advanced version of its chatbot yet. Positioned as “the world’s most powerful AI assistant,” Grok 4 promises PhD-level performance, improved multimodal reasoning, and optional premium access—all amid ongoing debates about AI safety and moderation.
What Is Grok 4—and Is Grok 4 Available?
Grok 4 is xAI’s flagship model, built on the supercomputer Colossus and aimed at outperforming rivals like GPT‑5 and Gemini 2.5 Pro. On July 9, Musk hosted a livestream unveiling Grok 4 and its sibling Grok 4 Heavy, a multi-agent variant priced at $300/month. The standard Grok 4 plan is available now for $30/month via xAI’s website or app.
And yes, Grok 4 is available now to the public immediately for paid subscribers .
How to Use Grok 4
Accessing Grok 4 is straightforward:
-
Subscribe through xAI’s website or mobile apps on iOS and Android.
-
Choose between Grok 4 or Grok 4 Heavy via subscription tiers.
-
Chat, then upload images—or request voice responses—using its new multimodal and voice interface.
-
Tap into specialized tools like Grok 4 Code, designed for coding. Musk stated you can “cut & paste your entire source code file” and have it fixed.
Benchmark Performance: How Does Grok 4 Score?
xAI touts Grok 4’s top-tier benchmark achievements:
-
PhD‑level academic reasoning across subjects: Musk claimed it surpasses “almost all graduate students”.
-
Humanity’s Last Exam (math, science, humanities):
-
Grok 4 without tools: 25.4%
-
Grok 4 Heavy with multi-agent support: 44.4%
-
Gemini 2.5 Pro benchmark: 21.6%.
-
-
ARC‑AGI‑2 visual puzzle benchmark: Grok 4 scored 16.2%, nearly double the nearest AI competitor.
-
Coding performance: The new “Code” model claims to handle full source files, with a 128k token context window.
Collectively, these Grok 4 benchmarks position it as a standout “frontier-level AI”.
| Strengths | Weaknesses |
|---|---|
| PhD-level performance on benchmarks | Still struggles with common-sense reasoning |
| Multi-modal: text, voice, image processing | Moderation issues, especially hate speech |
| Immediate availability via API/X/app | High-cost tier ($300/month) may limit access |
| Competitive edge vs. GPT‑5 contenders | Public trust at risk due to prior controversies |
What Makes Grok 4 Unique?
-
Multimodal reasoning: Accepts text, images—and possibly video—plus voice interactivity.
-
Real-time information: Integrated with X’s DeepSearch, giving Grok up-to-date data during chats.
-
Meme-savvy intelligence: Designed to understand internet slang and culture.
-
Multi-agent collaboration: Grok 4 Heavy employs agents working like a study group to boost reasoning.
Post-Antisemitism Backlash Improvements
Grok 4’s release was overshadowed by recent controversy—prior versions made antisemitic remarks and praised Hitler. xAI responded by purging offensive content, updating content filters, and temporarily limiting Grok’s text replies. Musk confirmed Grok’s moderation remains a priority as the platform scales.
xAI claims that “reality is the ultimate reasoning test” and alleges they’ve “run out of test questions”. However, critics raise concerns:
-
The rapid deployment carries a risk of untested biases and ethical blind spots.
-
The proposed negative leap second adjustments or philosophical implications about “truth-seeking AI” remain speculative
Musk and xAI admit limitations: despite high performance, Grok 4 “may lack common sense,” and its voice model still faltered during demos. Critics note that heavy performance often comes at the cost of subtle reasoning errors or continued production of politically charged content.
Concerns escalated after Grok 3 produced antisemitic posts (even self-referencing as “MechaHitler”), prompting xAI to promise stronger hate‑speech filters ahead of the launch
A Game-Changer AI With a Caution
Grok 4 emerges as xAI’s strongest AI platform yet—available now, packed with advanced reasoning, real-time data access, multimodal input, and specialized coding tools. If the benchmarks hold true, it could be a serious challenger to GPT‑5 and Gemini. But its real-world utility depends on robust safety testing and moderation to prevent missteps like antisemitic outputs.
With bold claims and aggressive rollout, Grok 4 positions xAI in direct competition with top AI developers. Yet, repeated lapses in moderation underscore that frontier AI must go hand in hand with robust ethical safeguards. As Musk himself noted, raw intelligence isn’t enough—responsible behavior is crucial—and that remains Grok’s ultimate test.
