.png)
November 20, 2024
Development in the field of Artificial Intelligence (AI) is progressing rapidly. xAI recently introduced the latest versions of its large language models: Grok-2 and Grok-2 mini. These models are designed to set new standards in chat, programming, and logical reasoning, and offer significant improvements compared to the predecessor model, Grok-1.5.
Grok-2 was tested under the pseudonym "sus-column-r" on the LMSYS chatbot arena, a well-known benchmark platform for language models, and surpassed models like Claude and GPT-4 in terms of overall rating. Internal tests by xAI, in which AI tutors evaluated the models in realistic scenarios, also confirm the progress of Grok-2, particularly in the areas of instruction following and providing accurate information.
The performance of Grok-2 and Grok-2 mini was evaluated using various academic benchmarks in areas such as logical reasoning, reading comprehension, mathematics, science, and programming. Both models show significant improvements over Grok-1.5 and achieve competitiveness with other leading models in areas such as graduate-level scientific knowledge (GPQA), general knowledge (MMLU, MMLU-Pro), and mathematical tasks (MATH). Grok-2 excels particularly in image-based tasks, achieving outstanding results in visual mathematics (MathVista) and document-based question answering (DocVQA).
xAI is integrating Grok-2 and Grok-2 mini into the X platform. Premium and Premium+ users have access to both models. Grok-2 offers advanced capabilities in text and image understanding and integrates real-time information from X. Grok-2 mini offers a balance between speed and response quality. The user interface has been redesigned and expanded with new features. xAI is also experimenting with the FLUX.1 model from Black Forest Labs to further expand Grok's capabilities on X.
xAI plans to release Grok-2 and Grok-2 mini for developers through a new enterprise API platform. The API is based on new technology that enables worldwide inference deployments with low latency. In addition, advanced security features such as mandatory multi-factor authentication, detailed traffic statistics, and billing analytics are offered. A management API enables the integration of team, user, and billing management into existing tools and services.
xAI plans to integrate Grok-2 into various AI-powered features on X, such as improved search functions, deeper insights into X posts, and optimized reply functions. A preview of multimodal understanding as a core component of Grok on X and the API is expected to be released soon. xAI continues to focus on advancing core competencies in the area of logical reasoning and plans to announce further developments in the coming months.
Bibliography: https://x.ai/blog/grok-2