I’ve had the Rabbit R1 for about three weeks, and it keeps putting me in the same spot. I pick it up to check something, and twenty minutes later I realize I’ve been talking to my agents through it like it’s the most natural thing in the world. Then I leave the house, and I remember why it’s not my daily driver.
The Rabbit R1 is $199 and about the size of half an iPhone. It’s plastic, but it doesn’t feel cheap. The box has the device and a clear case, nothing else. No cable, no paperwork. You’ll want the case, because there’s no bezel protecting that screen, and you’ll need a USB-C cable you already own.
Judged on hardware alone, it’s fine. The screen is readable in direct sun, in a dark room, everywhere in between. Touch feels like a modern phone, just barely slower, which is an 8.5 in my book. The scroll wheel is stiff enough that you won’t click it by accident, and the side button is easy to find by feel. The speaker gets genuinely unpleasant past 80 percent volume, so I keep it below that.
The mic is the surprise. It picks me up from arm’s length with no problem, and it filters out background noise impressively. Dogs barking, a laptop playing audio in the same room, it still hears me fine.
The verdict
Rabbit R1
A $199 AI-native pocket companion with camera vision, translation, and third-party agent support.
- Price
- $199
- Tested on
- Home WiFi, ~1.5 weeks
- Battery
- Half to 3/4 day, daily charge
- Connectivity
- WiFi + nano-SIM
- Big hook
- Voice gateway to your AI agents
What works
- Five-minute setup: account, link, done. Zero friction.
- Camera vision is genuinely good: object ID, text reading, context clues.
- Mic and noise handling: hears me through dogs barking and background audio.
- Push-to-talk voice gateway: steers Hermes, Claude, and OpenClaw hands-free.
- Modern-phone touch feel: 8.5/10 on responsiveness.
Tradeoffs
- False memory claim: said it would remember our conversation, then forgot it 30 minutes later.
- 10-20 second latency: every response waits. The biggest UX red flag.
- No wake word: push-to-talk only, walkie-talkie style.
- Phone-redundant on the go: needs your phone's hotspot to travel.
- Battery: half to three-quarters of a day, daily charging.
Tony's take
AI pocket companion
Rabbit R1
A $199 AI-native device with a 360-degree camera, push-to-talk voice, and third-party agent support. Designed with Teenage Engineering.
- Price
- $199
- Connectivity
- WiFi + nano-SIM
- Battery
- Half to 3/4 day
- Camera
- 360-degree rotating
- Designed with
- Teenage Engineering
Setup Takes Five Minutes
This is the most polished part of the whole device. Account creation, email verification, linking the device, it all just worked in about five minutes with zero friction. WiFi pairing and reconnection have been stable on my home network the whole time.
Connecting it to Hermes took a bit of fiddling, but that was on my end, not Rabbit’s. Once the bridge was up, it’s been running smoothly.
The Daily Friction
There’s no wake word on the Rabbit R1. It’s push-to-talk: hold the button, talk, release. It has a walkie-talkie feel, which is fine once you get used to it, but if you were hoping for a “Hey Rabbit” experience, that’s not this device.
The bigger problem is latency. Every response takes ten to twenty seconds. That’s the single biggest red flag in the whole review. A quick question becomes a small wait, every single time, and it breaks the flow of what should feel like a conversation.
Battery is honest but unimpressive. Half to three quarters of a day with normal use, and I charge it daily. Fast charging keeps it from being a real problem, but it’s not a strength.
The Trust Problem
The most damaging moment of this whole review wasn’t a missing feature. I asked the R1 whether it keeps conversation context across a gap, whether it would remember the thread of our conversation. It said yes, confidently. It would remember.
I set it down for about thirty minutes, came back, and asked it what the conversation had been about.
“I don’t have any previous conversation to refer to,” it said.
That’s two problems in one exchange. First, there’s no session persistence; the context was gone. Second, and worse, the device fabricated a capability claim. It didn’t say “I don’t retain context” or “I’m not sure.” It told me it would remember, with the same confidence it uses for everything else.
It can tell me who made it and what model it runs, and it gets those right. But it can’t accurately describe its own limits. For a $199 voice assistant, that’s a trust failure, not a missing feature. You can’t calibrate your expectations against a device that overclaims its own abilities.
Worth noting: the Hermes bridge retains context across sessions. So this is a native model failure, not a platform limitation. Rabbit’s own brain has the problem.
Camera Vision Is the Real Native Feature
Point the camera at something and it sees. It identified my laptop correctly, down to the ten-key pad and NVIDIA stickers, and it read what was playing on screen. It’s good at reading text and generalizing when there’s no branding visible.
Translation via camera: Spanish was flawless, Chinese completely failed. Mixed bag there, but the object recognition is legitimately the one core feature that does something a phone isn’t better at.
The Phone-Redundant Middle
Spotify works. Ask for an album, an artist, a song, and it plays. The journal exists. Both are fine. Both are things your phone already does better. Cool that they’re there, not a reason to buy.
The Killer Feature: A Voice Remote for Your Agents
This is what makes the Rabbit R1 worth keeping. I use it as a voice gateway to Hermes. I talk to it, and my agent stack answers with real context and real persistence. That’s a different product from what Rabbit’s own model offers, and it’s the reason the device has a place on my desk.
It’s also where the review’s thesis lives. If you’re deep in AI workflows, if you’re running agents and want to steer them by voice from something the size of a deck of cards, the Rabbit R1 is genuinely good at that. OpenClaw, Claude, whatever backend you want to wire up, it works as a small, fast input device for your AI stack.
The Structural Problem
And now the part that keeps the score at four. On the go, the Rabbit R1 has no independent connectivity. It needs your phone’s hotspot, and the moment you’re tethered to your phone, you’re holding the device that does everything the R1 does.
I could throw a SIM card in it. Rabbit supports nano-SIMs and cellular data. I’m not going to, because paying for a second data line to serve a device that duplicates my phone is a bad deal. This is structural, not incidental. The R1 requires the very device it’s meant to augment, so its mobile use case collapses into the phone’s.
The only places it wins are home WiFi, hands-busy moments, and quick voice sessions with my agents. That’s a real niche. It’s just not a phone replacement, and it doesn’t pretend to be.
Who Should Buy the Rabbit R1
This is for people deep in AI coding workflows who want to control their agents by voice from a small, lightweight device. If you live in terminals and agent logs and you want a physical object to talk to your stack with, this is the closest thing to a purpose-built tool for that.
It’s not for normal folk. If you have an iPhone or Android with Siri or Google Assistant, your phone already does everything the R1 does, with a better screen, a better camera, and no extra device to carry.
The Rabbit R1 Verdict
Four out of five stars. It’s easy to set up, easy to use, and honestly fun to use. The false memory claim keeps it from five, because a device you can’t trust to describe its own abilities is a device you have to be careful with. The Hermes bridge is why it’s not a three.
If you run agents, this is a fun, useful companion that earns its spot. If you don’t, save the $199 and keep using the phone you already own.
