Elon Musk’s AI startup, xAI has officially launched Grok-2 and Grok-2 Mini in beta, with advancements from the previous Grok-1.5 model.
These new AI models are currently available exclusively to X Premium and Premium+ users. Designed for enhanced chat, coding, and reasoning tasks on the platform, the models come with newly introduced image generation feature.
xAI’s blog post read: “We are excited to release an early preview of Grok-2, a significant step forward from our previous model Grok-1.5, featuring frontier capabilities in chat, coding, and reasoning. At the same time, we are introducing Grok-2 Mini, a small but capable sibling of Grok-2.”
Grok-2 has been tested under the alias “sus-column-r” on the LMSYS leaderboard, where it has outperformed well-known models like Claude 3.5 Sonnet and GPT-4-Turbo in overall Elo score, revealing improvements in reasoning, tool use, and handling complex sequences of events.
Grok-2 and Grok-2 Mini were built with advancements across various academic benchmarks, including reasoning, reading comprehension, math, science, and coding. The models have helped in areas such as graduate-level science knowledge (GPQA), general knowledge (MMLU, MMLU-Pro), and math competition problems (MATH).
In vision-based tasks, Grok-2 has brought state-of-the-art performance in visual math reasoning (MathVista) and document-based question answering (DocVQA).
A great feature of Grok-2 is its ability to generate images directly on the X platform. This ability is powered by FLUX.1 from Black Forest Labs, according to text below sample image prompts. However, this feature has already led to some issues.
Early users have pointed out that Grok-2’s image generation lacks guardrails, particularly concerning the creation of images involving political figures. With the U.S. presidential election on the horizon, this raises talks about the possibility of misuse, including the spread of misinformation.
Given the history of AI-generated content contributing to misinformation, xAI may face pressure to implement restrictions on this feature. The lack of clear guidelines or metadata showing that images are AI-generated further complicates the issue, making it difficult to distinguish between genuine and AI-created content.
Away from that, xAI plans to make Grok-2 and Grok-2 Mini available to developers through a new enterprise API later this month. This API will support multi-region inference deployments for low-latency access worldwide and include enhanced security features such as mandatory multi-factor authentication and advanced billing analytics.
The API platform is designed to integrate seamlessly with existing in-house tools and services, offering developers strong faculties for managing team, user, and billing functions.
In addition to the API, xAI is preparing to roll out Grok-2’s AI-driven features on X, which will include enhanced search abilities, post analytics, and improved reply functions. These enhancements are expected to bring AI-powered replies to the platform, further integrating Grok-2 into the user experience on X.
xAI is also planning to release a preview of multimodal understanding as a core part of Grok’s experience on X and API. This development will likely extend Grok-2’s applications beyond text and image generation, bringing a more thorough AI interaction to the platform.
Since announcing Grok-1 in November 2023, xAI has been advancing rapidly with a small, highly skilled team. However, the challenges of Grok-2’s image generation feature and the moral implications of its use cannot be overlooked.
xAI is expected to share further developments later on as it continues to refine and expand Grok-2.
Author
