SAN FRANCISCO — xAI has announced the early preview release of Grok-2 along with its compact sibling, Grok-2 mini, marking a significant evolution from the previous Grok-1.5 model. These new models deliver frontier capabilities in chat, coding, and reasoning, while integrating real-time information from the 𝕏 platform. In competitive benchmarks, an early version of Grok-2—tested under the name “sus-column-r” on the LMSYS leaderboard—has outperformed established models including Claude 3.5 Sonnet and GPT-4-Turbo.

Grok-2 and Grok-2 mini represent a leap forward in AI performance. Internally, AI Tutors rigorously evaluated these models based on instruction following and factual accuracy. Grok-2 exhibits significant improvements in reasoning with retrieved content and tool use capabilities, such as identifying missing information and logically processing sequences of events. Comprehensive benchmarks across academic tasks—including graduate-level science, general knowledge, math competitions, coding, and even visual math reasoning—show that both models are competitive with some of the best in the industry.

In a move to further enhance user experience, Grok-2 is now accessible to 𝕏 Premium and Premium+ users. The upgraded Grok experience features a redesigned interface that not only improves access to real-time information but also enhances functionalities like enhanced search capabilities, deeper insights on 𝕏 posts, and improved reply functions. Grok-2 mini, although smaller in size, strikes an attractive balance between speed and answer quality, ensuring versatility for a broad range of tasks.

In addition to the consumer rollout on 𝕏, xAI is set to extend Grok-2 and Grok-2 mini to developers via a new enterprise API platform later this month. The API is built on a bespoke tech stack designed for low-latency, multi-region deployments and enhanced security features such as mandatory multi-factor authentication. Developers will also benefit from rich traffic statistics, advanced billing analytics, and seamless management integration for team and user administration.

Since the announcement of Grok-1 in November 2023, xAI has demonstrated rapid progress with its focused, high-talent team pushing the boundaries of AI development. Grok-2 positions xAI at the forefront of the industry, with plans to soon preview multimodal understanding as a core feature of the Grok experience on 𝕏 and its API. As xAI continues to drive innovation in AI reasoning capabilities and real-time data integration, the company is actively seeking new talent to join its mission of developing transformative AI technologies.

In the context of AI, what’s the radical shift here?

The radical shift is the integration of real-time, interactive data with significantly enhanced reasoning and tool-use capabilities. Grok-2 and Grok-2 mini leverage the 𝕏 platform to provide up-to-date, context-aware responses while setting new performance benchmarks across diverse tasks, from chat and coding to complex reasoning and vision-based applications. This marks a move from static, pre-trained models to dynamic, real-world informed AI that is both more capable and more adaptable in real time.