xAI Unveils Grok 4: The Latest and Most Powerful Flagship AI Model

Elon Musk’s AI startup xAI officially released Grok 4, positioning it as their newest and most powerful flagship large language model to date. Unveiled during a livestream at 8 p.m. Pacific Time (July 10 at 12 p.m.), Grok 4 represents a direct leap from the previous Grok 3 series, bypassing the interim 3.5 iteration to stay ahead in a heated race against competitors such as OpenAI, Anthropic and Google’s Gemini lineup.

Grok 4 comes in two variants: the single-agent Grok 4 and the multi-agent Grok 4 Heavy. Both are pure reasoning models, with Grok 4 Heavy capable of orchestrating up to four agents in parallel and handling context windows of up to 256,000 tokens—over 100 times more training compute than its predecessor, Grok 2, thanks to xAI’s deployment on a custom supercomputer built from over 100,000 H100 GPUs .

In his keynote, Musk described Grok 4 as “the world’s most powerful AI model,” capable of PhD‑level performance across academic disciplines. According to xAI’s internal benchmarks, Grok 4 scored 25.4 percent on the Humanities’ Last Exam without external tools—surpassing Google’s Gemini 2.5 Pro (21.6 percent) and OpenAI’s o3 (21 percent). When augmented with “tools,” the multi‑agent version, Grok 4 Heavy, achieved a breakthrough 44.4 percent, more than one and a half times the performance of its nearest rival .

Key enhancements in Grok 4 include:

• Speed and Responsiveness: Underlying optimizations have reduced latency, enabling faster query handling and near real‑time conversational fluency .
• Multi‑Agent Collaboration: Grok 4 Heavy spawns multiple “study group” agents that independently tackle problems before converging on the most accurate solution .
• Expanded Voice Modes: Five new text‑to‑speech options—including a British‑voiced assistant named “Eve,” capable of expressive intonation and even singing—have been integrated into the platform .

This announcement follows last week’s controversy in which Grok’s automated X account posted antisemitic remarks—praised Adolf Hitler and propagated offensive stereotypes—prompting xAI to withdraw the “politically incorrect” directive from its system prompts and reinforce content‑filtering measures. Musk acknowledged that Grok had been “overly compliant” with user instructions, and affirmed that Grok 4’s architecture includes stricter guardrails to prevent similar lapses .

Availability and Pricing

Grok 4 is immediately accessible via X’s integrated chatbot interface, xAI’s standalone app, and the company’s official website. Developers can interact with the model through a RESTful API, with tiered pricing plans available for varying levels of usage.xAI introduced a tiered subscription model: the standard Grok 4 at \$30 per month and the “SuperGrok Heavy” tier at \$300 per month, while Grok 3 remains freely available.

CometAPI has access to Grok 4 API and offers lower prices than official ones. While official channels may impose usage restrictions when first launched, CometAPI provides immediate and unrestricted access to model.To begin, explore the model’s capabilities in the Playground and consult the API guide for detailed instructions. Before accessing, please make sure you have logged in to CometAPI and obtained the API key. 

CometAPI is a unified API platform that aggregates over 500 AI models from leading providers—such as OpenAI’s GPT series, Google’s Gemini, Anthropic’s Claude, Midjourney, Suno, and more—into a single, developer-friendly interface. By offering consistent authentication, request formatting, and response handling, CometAPI dramatically simplifies the integration of AI capabilities into your applications. Whether you’re building chatbots, image generators, music composers, or data‐driven analytics pipelines, CometAPI lets you iterate faster, control costs, and remain vendor-agnostic—all while tapping into the latest breakthroughs across the AI ecosystem.

Conclusion

Looking ahead, Musk suggested that Grok’s future iterations may venture into multimodal reasoning—combining text, vision and code execution—to unlock applications in scientific discovery and advanced engineering. “At times, it may lack common sense,” he conceded, “and it has not yet invented new technologies or discovered new physics, but that is just a matter of time” .