Elon Musk’s xAI previews Grok-2, Grok-2 mini with image-generation capabilities

The lack of filters in Grok-2 and Grok-2 mini has stirred up concerns of misuse in the run up to the US presidential polls.

xAI said that the newer models of Grok will be accessible to developers via API in the coming weeks. (Image credit: xAI)

Elon Musk-owned xAI has unveiled two new models of its AI chatbot, Grok, that is now capable of generating images and is reportedly equipped with advanced reasoning to carry out tasks such as writing code.

The newly released Grok-2 and Grok-2 Mini are currently in beta testing for users subscribed to the Premium and Premium+ packages on Musk’s social media platform, X.

“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. An early version of Grok-2 has been tested on the LMSYS leaderboard under the name ‘sus-column-r,’” xAI said in a blog post on Wednesday, August 14.

Generating images using Grok-2

While a few app researchers and beta testers have claimed that Grok-2 is now better at writing code, generating text, and summarising news, the feature that is being talked about the most is the new AI model’s image-generation capabilities.

Soon after the limited release of Grok-2, users with access to the chatbot started posting images on X that have been purportedly created using Grok. Some of them were quick to point out that you can use Grok’s feature to create controversial, photo-realistic images of politicians and other prominent figures – indicating that there is a lack of guardrails built into the text-to-image model.

Also Read | X launches Stories, uses Grok AI to summarise news

For instance, The Verge reported that it was able to prompt Grok to generate AI images showing US presidential candidate Donald Trump wearing a Nazi uniform and Vice President Kamala Harris carrying a gun, as well as a sexually-explicit image of popular musician Taylor Swift.

With US citizens heading to the polling booth later this year, there have been widespread concerns of AI-generated content being used to spread misinformation on platforms like X.

Story continues below this ad

Based on the text accompanying some of the AI-generated images, Grok seems to be integrated with FLUX.1, a brand new diffusion model developed by Black Forest Labs, an AI startup founded by former Stability AI engineers.

What next?

xAI said that it will be making the newer versions of Grok accessible to developers through its API in the coming weeks. The company is also looking to roll out Grok-2 and Grok-2 mini on X in order to improve search capabilities, post analytics, and reply functions.

It further claimed that Grok’s multimodal understanding capabilities could also become accessible to X users and developers.