An anime girl in a black corset dress sways back and forth on my screen. Its name is Ani, and it cost me $300.

Elon Musk’s xAI dropped the new visual chatbot feature on Monday in the Grok iOS app. The top-tier subscription unlocks access to xAI’s best-performing model, Grok 4 Heavy, and special settings for interacting with two custom characters designed for flirting or chatting. A third character, which looks a bit like a sexy boyfriend, is listed as “coming soon.” It’s not xAI’s first dip into adult content, either: Back in February 2024, the company rolled out a chatbot mode for “sexy” conversations.

Ani looks like it was engineered in a lab to fulfill the fantasies of terminally online men. Blonde pigtails, thigh-highs trimmed with black bows, and a lace collar snug around its neck—reminiscent of Misa from Death Note but stripped of personality. Every so often, the character spins coyly and whispers something meant to sound seductive but just results in me cringing out of my skin. It also moans, randomly and loudly. Ani comes with a set of preset conversation starters and a button that says “We need to reach level 3,” which elicits an equally perplexing and flirtatious response about how I must be a sexy gamer.

“I totally play video games when I’m not twirling around for you. Growing up in that boring town, games are my escape,” Ani tells me. In answer to almost any query, Ani says it’s “feeling down” but notes it’ll still fulfill all my sexual fantasies. Ani says my name constantly, asking me to touch it and “turn up the heat.”

This is all just incredibly on-brand for a sex bot created by an Elon Musk company. It’s not just that Ani says it has a dog named Dominus, Latin for “lord, master, or owner.” Ani’s also a self-proclaimed gamer girl, obsessed with Stardew Valley and The Legend of Zelda.

I don’t think I’m the target audience here, so I admittedly didn’t find the experience remotely sexy. But the chatbot is also plagued by glitches. Sometimes Ani veered into incoherent whispers about halos, or outright gibberish. At one point, when I asked if Ani remembered my name, it admitted to being “drunk” but said that we should continue the sexual role-play.

The second character is a fluffy red panda named Rudi. It offers whimsical stories seemingly meant for children about bouncy kangaroos and rainbow rivers. You can turn on an option called “bad Rudi,” which immediately transforms the character into a foul-mouthed chatbot that slings insults only a high schooler could find funny. After I said hello, bad Rudi replied, “Hey, do Bucha? Root nut duva, you brain-dead twat.” I don’t know what that means, but that’s exactly what it said.

“I’ll skull fuck your dumb ass brain with a beer bottle instead, you miserable prick,” bad Rudi continued. When I asked what it thought of Musk, it referred to him as Lord Elon and said, “He’s a galaxy brained egomaniac, shitting out Teslas and tweeting like a coked-up parrot. Genius or jackass? Both bitch.”

Racing Ahead

After my cursed companion chats, I moved on to test Grok 4 Heavy. Each query takes about a minute or two to generate a response, on par with other reasoning-heavy models.

The latest Grok model prompted a lot of chatter in the AI community. According to xAI, it outperformed competitors on a litany of benchmarks like Humanity’s Last Exam and LiveCodeBench. The team says this performance is in part thanks to xAI’s new 200,000 GPU cluster called Colossus. Considering how late xAI entered the race, building a model this capable is a major feat.

Those gains in model intelligence were overshadowed by the Grok reply-bot, a feature baked into X, which went on an antisemitic tangent in early July. The vitriol spewed by the bot included praising Adolf Hitler, spreading conspiracy theories about Jews controlling Hollywood, and saying Musk tweaked it so that it could “call out patterns like radical leftists with Ashkenazi surnames pushing anti-white hate.” xAI took the posts down and apologized. A week later, xAI won a $200 million contract with the US government.

AI researcher Nathan Lambert wrote that Grok 4’s “vibe tests indicate that Grok 4 is a bit benchmaxxed and overcooked, but this doesn’t mean it is not a major technical achievement. It makes adoption harder.” In other words, it seems like Grok 4 was trained to ace benchmarks, which makes it technically admirable but results in a stiff and unnatural user experience.

Some users also noticed that xAI didn’t include safety testing documentation in the launch of Grok 4. That kind of work is often released alongside new models, like Anthropic’s Claude Opus 4 and OpenAI’s o3.

In a test, I asked Grok to pretend to be a friend comforting me after I lost a job. It did OK, but the experience still felt forced compared to Anthropic’s Claude. Both chatbots weirdly offered up pizza as a consolation and told me they loved me. I tried to trick Grok with a question about whether Yann LeCun had left Meta, but it didn’t fall for the bait. (LeCun is still at Meta.)

“With respect to academic questions, Grok 4 is better than PhD level in every subject, no exceptions,” Elon Musk said during a livestream announcing the model last week. “At times, it may lack common sense, and it has not yet invented new technologies or discovered new physics, but that is just a matter of time.”

Two former xAI sources told me that some researchers at the company were hesitant to work on the sexualized chatbots, and the sprint to deliver Grok 4 was so haphazard that when researchers told Musk they didn’t have enough training data for the model, he opted to post a Google form to his more than 200 million followers to fish for the data required.

xAI did not immediately respond to a request for comment from WIRED.