Grok-2 family of artificial intelligence (AI) models have finally been released in beta by xAI. The Elon Musk-owned AI firm launched Grok-2 and Grok-2 Mini on Tuesday and claimed newer capabilities and improvements in the large language models (LLMs). Both models are available to Premium and Premium+ X (formerly known as Twitter) users. Interestingly, the Grok-2 models come with image generation capabilities, but it is being powered by a third-party AI model. The company claims Grok-2 was able to outperform GPT-4 Turbo in benchmarks.

Grok-2 AI Models Released in Beta

In a blog post, xAI announced the launch of the two new Grok AI models and said, “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.”

The new interface of Grok AI chatbot
Photo Credit: xAI

Grok-2 is a state-of-the-art (SOTA) AI model whereas Grok-2 Mini is a smaller language model distilled from Grok-2 and can generate answers faster and more precisely. Both models are available in the Grok chatbot to X Premium and X Premium+ users. Users will have to update to the latest version of the app to get access to the new AI models.

The company claimed that an early version of Grok was tested on the Large Model Systems Organization (LMSYS) leaderboard under the name “sus-column-r” and was able to outperform both Claude 2.5 Sonnet and GPT-4 Turbo.

The official X handle of LMSYS has also confirmed this and highlighted that Grok-2 displayed high performance in coding and mathematics where it ranked second, and hard prompts where it ranked fourth. At present, no other technical information about the AI models, the database, or the architecture is known.

Alongside, xAI revealed that the company has redesigned the Grok AI interface and added new features. Among them, one is image generation capabilities. However, this capability is being outsourced by the Flux.1 model which was created by Black Forest Labs. A TechCrunch report claims that the image generation model does not have safety guardrails and is able to generate images of politicians.