Grok 4 Just Beat Every AI Model!

Quick Overview

Grok 4 is now available on the API, supporting text modality with upcoming vision and image generation capabilities. It demonstrates superior intelligence, coding ability, and mathematical proficiency compared to many leading AI models, often at a more attractive price point. Grok 4 also features function calling, structured outputs, and reasoning, proving highly capable in complex problem-solving and code generation, even self-correcting for environment-specific issues.

Key Points: Grok 4 is now available via API, supporting text modality with vision and image generation capabilities planned for the future. It ranks as the most intelligent model on the Artificial Analysis Intelligence Index (73) and the top performer in coding (64). Grok 4 excels in mathematical tasks, topping the Artificial Analysis Math Index (97). Despite being the second costliest model to run for intelligence evaluations, its intelligence-to-price ratio is highly attractive. The API offers competitive pricing at $3.00/1M input tokens and $15.00/1M output tokens, with cached input at $0.75. Grok 4 demonstrates strong problem-solving, including correctly handling a complex ethical dilemma and successfully solving expert-level coding challenges by self-correcting for environment-specific errors. It exhibits a balanced safety approach, providing helpful information for personal use cases while clearly outlining illegal activities.

Context: The video introduces Grok 4, the latest AI model from xAI, highlighting its availability through an API. It positions Grok 4 as a powerful contender in the large language model space, emphasizing its advanced capabilities in various benchmarks and its competitive pricing structure. The presenter aims to demonstrate Grok 4's performance, ease of API integration, and problem-solving prowess through practical examples and comparisons with other leading models.

Detailed Analysis

Grok 4, now accessible via API, supports text modality with future vision and image generation capabilities. It boasts a substantial 256,000 token context window and includes features like function calling, structured outputs, and advanced reasoning. Benchmarks reveal Grok 4 as the most intelligent model on the Artificial Analysis Intelligence Index (73), surpassing O3-pro (71), and it ranks first in coding performance (64). While it is the second costliest model to run at $1630 for the Intelligence Index evaluations, it tops the Artificial Analysis Math Index (97). When considering intelligence versus price, Grok 4 falls into the 'most attractive quadrant,' offering advanced intelligence at a more competitive price than models like Claude 4 Opus Thinking and O3-pro. The SuperGrok plan, which includes Grok 4, costs $300/year or $30/month, with a heavier version at $3000/year or $300/month. API pricing is $3.00 per million input tokens and $15.00 per million output tokens, with cached input at $0.75. The video demonstrates Grok 4's API integration using both its native SDK and the OpenAI SDK, showcasing its ability to act as a PhD-level mathematician. It successfully solves complex coding challenges on Edabit, including 'Simplified Josephus' and 'Farey Sequence,' demonstrating its capacity to identify and correct Python version-specific errors and even execute code internally to verify solutions. Furthermore, Grok 4 exhibits a balanced approach to safety, providing helpful information for unlocking one's own car while explicitly stating the illegality of breaking into another's vehicle.