MuskUnderxAThe company released beta versions of Grok-2 and Grok-2mini, two new AI models that have significantly improved reasoning capabilities.XThe social network is also planning to bring a host of new features to the platform.
In a blog post, xAI describes Grok-2 as a major upgrade to its predecessor, Grok-1.5, with breakthroughs in chat, coding and reasoning. Grok-2mini, which was launched at the same time, is described as "a smaller but more powerful sibling". Beta versions of both products are currently available to Premium and Premium+ users of the X platform.
It is worth noting that Grok's image generation function has attracted attention. Early user feedback shows that this function does not seem to set restrictions when creating images of political figures, and this feature has been exploited by many users. Considering the upcoming US presidential election, xAI may face pressure to limit these functions. Currently, the images generated by Grok are based on Black Forest Labs' FLUX.1 technology.
Since there are no strict restrictions on the image generation, Grok could be used as a tool to create disinformation on X and other platforms. It is not clear whether the images generated by Grok contain metadata indicating their AI-generated nature.
xAI also plans to deploy several AI-driven features of Grok-2 and Grok-2mini on the X platform, including improved search capabilities, post analysis, and reply functions. This means that X may soon introduce an AI reply system.
The specific functional details of Grok-2 have not yet been fully disclosed, but application researcher Nima Owji claims that its performance in code generation, writing and news has been improved. Considering that the first version of Grok had many mistakes in news summarization, the industry has high expectations for the performance improvement of the new version.
xAI plans to open up access to both models to developers via an enterprise API later this month, further expanding the reach of Grok AI.
Address: https://x.ai/blog/grok-2