Released the important update Grok-2, which provides improved conversational, coding and reasoning skills.
In addition to Grok-2, xAI has also released Grok-2 mini, a scaled-down but functional variant of the primary model. Both will be accessible through xAI's enterprise API later this month and are already in beta on X.
xAI claims to outperform OpenAI's GPT-4-Turbo and Anthropic's Claude 3.5 Sonnet at the time of the announcement. However, it is important to note that GPT-4o currently leads in terms of overall AI assistant skills, with Google's Gemini 1.5 in second place.
AI Tutors are used in xAI's internal evaluation process to evaluate models on a range of real-world activities. "Grok-2 has shown remarkable improvements in its ability to reason with retrieved content and in its tooling capabilities, including accurately recognizing missing information, reasoning through sequences of events and eliminating irrelevant messages," claims the company.
According to benchmark data released by xAI, Grok-2 and Grok-2 tiny both show significant improvements over Grok-1.5. The models show competitive performance in domains such as general knowledge, graduate-level scientific knowledge and math competition tasks. Especially in vision-based activities, Grok-2 performs exceptionally well, with advanced skills in document-based question answering and visual mathematical reasoning.
There are new features and an updated UI in the Grok experience on X. Grok-2 and Grok-2 mini will be available for Premium and Premium+ customers. According to xAI, Grok-2 is "more intuitive, manageable and versatile for a wide range of tasks, whether you're solving coding problems, searching for answers or collaborating on writing projects."
To further expand the capabilities of Grok on X, xAI is also working with Black Forest Labs to test their FLUX.1 model.
Later this month, xAI debuts an enterprise API platform for developers. It will offer comprehensive analytics for billing, rich traffic information and enhanced security measures. To incorporate team, user and billing management into the tools and services currently in use, a management API will also be made available.
In the future, multimodal understanding will be a fundamental feature of the Grok experience on both X and the API, according to xAI. The reason for its rapid development since the unveiling of Grok-1 in November 2023 is "a small team with the highest density of talent."
To be at the forefront of AI development, xAI's new computational cluster is focused on improving basic reasoning skills. But the company has decided to stop using specific EU data to build its models.
Even though the introduction of Grok-2 is a big step forward for xAI, it is clear that there is still fierce competition in the AI market. The battle for AI supremacy is far from over, with ChatGPT-4o and Google's Gemini 1.5 leading the way and major companies like Anthropic still making rapid progress.
