xAI launches Grok-2: The AI assistant that surpasses GPT-4 and Claude in reasoning

Grok-2, xAI's new artificial intelligence model, marks a milestone in the development of virtual assistants. Launched in beta version on August 13, 2024, this advanced language model demonstrates superior capabilities in chat, coding, and reasoning, surpassing renowned competitors such as GPT-4-Turbo and Claude 3.5 Sonnet in key benchmarks.

The launch includes two variants: Grok-2 and Grok-2 mini, the latter designed to offer a balance between speed and response quality. Both models are available to Premium and Premium+ users on the X platform, formerly known as Twitter, through a redesigned interface in the app.

Grok-2 particularly excels in tasks requiring advanced reasoning, such as understanding retrieved content and using tools. In internal evaluations, the model has shown significant improvements in identifying missing information, reasoning about event sequences, and discarding irrelevant information.

Regarding academic benchmarks, Grok-2 and Grok-2 mini stand out in areas such as graduate-level scientific knowledge (GPQA), general knowledge (MMLU, MMLU-Pro), and competition math problems (MATH). Additionally, Grok-2 achieves state-of-the-art performance in visual mathematical reasoning (MathVista) and document-based question answering (DocVQA). xAI plans to launch an enterprise API for Grok-2 and Grok-2 mini by the end of August, offering advanced features such as multi-factor authentication, detailed traffic statistics, and billing analysis. This API will allow developers to integrate Grok's capabilities into their own applications and services.

The xAI team, known for its high density of talent, promises more innovations in the coming months, including a preview of multimodal understanding as a central part of the Grok experience on X and API. With this launch, xAI positions itself at the forefront of AI development, challenging tech giants and promising significant advances in the future of artificial intelligence.