Elon Musk's AI model Grok 2 is here and can generate images with political motifs
Elon Musk's AI company xAI has unveiled its new Grok 2 language model and a smaller Grok 2 Mini variant.
Initial tests indicate the models are competitive with other top AI systems in various benchmarks. Grok 2 also appears to have fewer restrictions on image generation compared to other providers. Both models are now available for X users.
According to xAI, both Grok 2 and Grok 2 Mini show significant improvements over Grok 1.5 on academic benchmarks in reasoning, reading comprehension, math, science, and programming. They perform on par with SOTA models in graduate level science (GPQA), general knowledge (MMLU, MMLU-Pro), and math competition tasks (MATH).
On the LMSYS ranking, a popular comparison benchmark for language models, an early version of Grok 2 called "sus-column-r" ranked third overall, matching GPT-4o's score. The model excels in programming (2nd place), difficult prompts (4th place), and mathematics (2nd place).
Early user reports suggest that Grok 2's image generation has fewer limitations than other providers. For example, it can generate political illustrations and images of real people, while ChatGPT and other image generators mostly refuse such requests, or at least try to.
One user noted Grok 2 seems unconcerned with copyrights. Another user said Grok 2's image generation outperforms Meta Llama's without "dumb guardrails."
The ability to generate images of real people, especially politicians, and AI propaganda raises concerns about misinformation, especially in the run-up to the US elections. Most major model providers block such content to prevent fake political images.
Another risk of AI-generated fake images is that people will claim that real images are AI-generated to discredit them. Essentially, this leads people to stop trusting everything they hear, read, and see unless it comes from a source they trust, making them easy prey for those in authority. Donald Trump has claimed at least twice that real images are AI-generated, fueling conspiracy theories.
Recently, U.S. secretaries of state sent Musk an open letter asking about AI-generated fake news on X. Musk himself shared a video with an AI-cloned voice of Vice President Kamala Harris calling Joe Biden "senile."
xAI works with Black Forest Labs to expand Grok's image capabilities on X by testing the FLUX.1 model. Premium subscribers can try Grok 2 by updating the X app. A new Enterprise API for developers is coming later this month.
Musk claims Grok 3, expected late this year, will be the most powerful AI system ever created.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe nowAI news without the hype
Curated by humans.
- Over 20 percent launch discount.
- Read without distractions – no Google ads.
- Access to comments and community discussions.
- Weekly AI newsletter.
- 6 times a year: “AI Radar” – deep dives on key AI topics.
- Up to 25 % off on KI Pro online events.
- Access to our full ten-year archive.
- Get the latest AI news from The Decoder.