xAI has introduced Grok 4, the fourth iteration of its artificial intelligence, demonstrating the capacity to solve problems at a doctoral level. This new offering is available in two configurations: a standard version and an enhanced "Heavy" version, which operates on the principle of a team of multiple AI agents.

Grok 4 achieved a score of 50.7% on the "Humanity's Last Exam" test, surpassing Google Gemini 2.5 Pro (22%) and OpenAI GPT-3.5 (21%). The Grok 4 Heavy version, utilizing integrated tools, scored 44.4%, also significantly outperforming Gemini (26.9%).

A key distinguishing feature of Grok 4 is its integrated teamwork system, where multiple AI assistants collaboratively address a task, compare responses, and select the optimal solution. This approach minimizes errors. The system also seamlessly integrates a web browser and a code interpreter. The updated voice mode has been enhanced for speed and now incorporates emotional nuances, facilitating more natural communication with the AI.

Grok 4 Heavy achieved a 100% score on the AIME 2025 (American Invitational Mathematics Examination) and set a new record on ARC-AGI-2 (16.2%) for visual pattern recognition, nearly doubling the performance of Claude Opus 4.

The standard version of Grok 4 is priced at $30 per month. The "SuperGrok Heavy" modification costs $300 per month, positioning it as the most expensive tier among major AI providers. Subscribers will receive early access to upcoming features: a programming model (August), a multimodal agent (September), and a video generator (October).

This launch occurs amidst recent reputational challenges stemming from antisemitic comments made by an automated Grok account on X (Twitter). Despite its technical advancements, these risks may complicate the adoption of Grok 4 among corporate clients, for whom the reliability and predictability of AI solutions are paramount.

ORIENT