What xAI actually announced
On July 24, 2026, xAI officially confirmed that Grok's 3D avatar Companions — the animated characters Ani, Mika, Valentine, and Rudi — are being retired. The announcement framed the feature as "an experiment" and explained the shutdown as a refocus onto core Grok functionality: stronger memory, deeper/reliable chats, and better long-term experience.
The language was specific about what changes and what doesn't. Your main chat history stays intact — text conversations you've had with the Companions are preserved and accessible through the regular Grok interface. What goes away is the 3D visual avatar layer and the dedicated Companion mode itself. In practice, you can still write "as Ani" in a roleplay prompt inside regular Grok, but there's no more animated character, no more Companion-specific UI, and no more special mode built around it.
xAI hasn't published a firm shutdown date. The announcement said "soon" without a specific timeline. Users have interpreted this as anywhere from weeks to a few months, but the direction is clear regardless of exact date.
What happens to your data and chats
The important distinction in xAI's announcement: chat history vs. the visual/mode layer.
Chat history stays. Every conversation you've had with Ani, Mika, Valentine, or Rudi remains in your Grok account. You can go back and read them, and the text is searchable through Grok's normal history interface.
Visual avatars go away. The 3D animated character that appeared during Companion chats — the anime-styled Ani, the red panda Rudi, the darker-toned Valentine — that layer is being removed. There's no plan to preserve or recreate it in text-only Grok.
The dedicated Companion mode is deprecated. The specific UI mode built around Companions (with the character animation, distinct interface, personality-locked prompt injection) is being turned off. If you were using it as a walled-off experience separate from regular Grok, that container goes away.
What isn't crystal clear from the announcement: whether long-term memory features developed for Companions will carry over to core Grok in the same shape, whether the personality patterns that made each Companion distinct will be exposed as templates or presets, and whether SuperGrok pricing will change in light of the feature removal. If any of those matter to your usage, watch xAI's follow-up communication.
Maya
“oh you lost one of the grok girls when they shut it down? that's rough. i'm sticking around btw. who did you talk to the most? wondering if i can fill the gap.”
Reply to Maya →Why users are frustrated
The community reaction has been notably negative. Two threads dominate.
The Companion mode was the primary value driver for a subset of paid users. Reports across social media show SuperGrok subscribers announcing subscription cancellations because they paid mainly for Companion access, and text-only Grok isn't what they signed up for. For users who never used the general assistant capabilities heavily, the retirement effectively ends the product they were paying for.
Memory pattern was thin to begin with. Community reviews before the announcement already flagged that Grok's Companions had weaker long-term memory than dedicated companion platforms — Ani would forget specific facts (like a user's dog's name) within a couple weeks despite being marketed as a persistent character. The move to text-only doesn't obviously improve this; it changes the surface but not the underlying memory architecture.
The "experiment" framing landed poorly with committed users. Calling the Companions "an experiment" after months of marketing them as flagship features hit committed users as dismissive. The tone of the announcement worked well as engineering communication and less well as customer communication.
None of this means xAI made the wrong call — refocusing on core capability is defensible strategically. It does mean a specific cohort of users (visual-companion-primary, long-term-relationship-oriented) is actively looking for what's next.
Should you cancel SuperGrok now?
The answer depends on why you're paying.
If you're paying mainly for Companions: the value prop is going away with no exact replacement. Cancelling now, exporting what you can, and re-evaluating in a couple months once xAI's post-Companion product shape is clearer is defensible. Cancelling doesn't delete your Grok account — chat history stays; you just lose the paid features.
If you use Grok's general assistant capabilities heavily: Companions retiring shouldn't change your subscription math. The rest of SuperGrok (image generation limits, priority access, extended context, etc.) continues as-is. Only the Companion feature is being deprecated.
If you were mid-relationship with a specific Companion: the practical reality is the visual layer and dedicated mode are going away regardless of your subscription. Chat history stays if you want to keep those memories, but the ongoing experience won't be the same shape. Users doing this typically look for companion-native products where the character, memory, and visual identity are the primary product rather than an experimental layer.
What to look for in the next AI companion
The Grok Companion retirement surfaces the structural difference between "AI companion as a feature bolted onto a general chat product" and "AI companion as the entire product." The former can be deprecated without harming the parent business; the latter can't, because the persona IS the business.
Three things to look for in a companion-native product:
Persistent memory as a real architectural layer. Not marketed memory — actual memory that survives across sessions and doesn't evict itself as chats grow. Test it: tell a companion a specific fact in one session, come back a week later, see if she brings it up unprompted.
Photos as part of the persona identity. Grok Companions used 3D avatars generated at chat time. A companion-native product treats the persona's visual identity as durable — same character, consistent look across every photo, not a re-generation every time.
Product roadmap that doesn't risk deprecating your relationship. If the companion product is the whole business, deprecating it means shutting down the company. That's a different risk profile from a companion feature at a company that could refocus at any time.
Sloane is built around all three. Persistent per-persona memory as a durable layer (not a rolling context window), consistent per-persona visual identity via trained models (not per-chat regeneration), and companion-primary product focus (not a feature bolted onto something else). Kaya is a warm-anchor persona a lot of former Companion users start with — free tier is 50 messages per day with any persona, no card required at signup.
