The Rabbit R1 at 6 Months: Surprisingly Smarter, but Still Overshadowed by Our Phones

With the Rabbit R1, tech startup Rabbit Inc. promised a peek at a future in which we don’t need to rely on our phones as much to get things done. But when the AI-powered handheld device launched in April, it didn’t come close to delivering on that pledge.

Six months later, I still find myself reaching for my phone instead of the R1. But I can’t deny that this tiny, palm-size gadget has vastly improved since I initially tested it six months ago. Longer battery life, faster response times and easier navigation gestures make the Rabbit R1 far more usable than it was at launch.

But these fixes don’t address my larger concern: Do you really need the Rabbit R1? Even with these improvements, using your phone is still easier and more intuitive. The R1 has gotten a lot smarter in recent months and can handle complex, multistep questions, making it seem far ahead of the virtual assistants we’ve been using for the past decade. 

The problem, however, is that phone-based virtual assistants are progressing equally fast, if not faster. Rabbit founder Jesse Lyu is trying to sell us on the promise that an AI-first interface is more efficient and natural. But phone makers have the same idea and their devices are more functional and useful than the R1, plus they’ve already earned a place in our lives. 

Lyu’s vision is on the right track when it comes to where mobile software could be headed. Making mobile interfaces more intelligent so that you don’t have to spend time jumping between apps is an ambition that tech giants like Google, Qualcomm and Apple all seem to be working towards in their own ways. 

I just don’t think the R1 will be the device we choose to use these interfaces on. 

Read more: I’ve Been Using Apple Intelligence For Weeks, and One Feature Stands Out

What I like about the Rabbit R1 six months later

First, the good: The Rabbit R1’s performance has undoubtedly improved, thanks to a wave of updates in the last six months. In addition to quality of life improvements, such as making it easier to access the settings menu using the scroll wheel, the R1 has gotten better at its most important job of answering questions.

When I asked how long it would take to get from Astoria, Queens to Greenpoint, Brooklyn at around 6 p.m. and whether it’s faster to drive or take the subway, the R1 gave me an accurate and helpful answer. It said it would take about 20 minutes by car and 30 minutes by subway, and provided high-level directions in bullet points that lined up with Google Maps‘ recommendations. It also provided a conclusion saying that driving is faster by about 10 minutes, but that I should consider traffic and parking if I decide to drive.

I’ve also found the Rabbit R1 to be helpful for time management and schedule planning. I asked the Rabbit R1 what time I should wake up in the morning if I need to be in the shower by 7:40 a.m. and want to squeeze in a workout beforehand. I also told it that I usually need 15 minutes to stretch and warm up first. Rabbit responded by advising that I wake up at 6:55 a.m. so that I could spend 15 minutes stretching before a 30 minute workout, which aligns with my usual routine.

That would be more impressive if it wasn’t for the fact that Google’s Gemini helper, which is available for free on Android phones and the iPhone, provided the same answer. It yet again underscores the point that you probably don’t need anything other than your phone to get the benefits of a smarter, more conversational assistant. Still, it is a big step forward for the R1, considering my experience with it in April was less reliable, more glitchy and far less capable.

Read more: I’m Not Impressed With AI on Phones Yet. What It Will Take to Change My Mind 

Even so, the R1’s responses aren’t perfect. Like many AI chatbots and voice assistants, it sometimes produces answers that are incorrect or don’t line up with my prompt. For example, I told the R1 that I like drama TV shows with a sci-fi twist, but I don’t want to watch anything that’s too long. Then I asked it to recommend shows for me that are streaming on Max or Netflix. But its suggestions included a show that wasn’t on either streaming service (Silo, which is on Apple TV Plus), and a show that doesn’t involve sci-fi elements (Everything Now, which IMDB classifies as a comedy drama). 

Gemini, on the other hand, gave me stronger recommendations like Stranger Things, The Umbrella Academy, Black Mirror and Westworld. However, its answers weren’t flawless either. It suggested The Time Traveler’s Wife, which is a film rather than a TV show. Plus, several of the series it recommended are quite long (i.e., Westworld, Stranger Things) even though I asked for shorter shows. It’s further evidence that AI-generated answers aren’t foolproof and still need to be checked, whether you’re searching on a phone or on a device like the Rabbit R1.

But the Rabbit R1 gave one answer that truly wowed me. I launched the camera, pointed it at the inside of my fridge, and asked it what I can make with the ingredients it sees. Google was able to provide some recipes too, but Rabbit’s recommendations were a little more specific. 

For example, it noticed that I had grapes, pickles, eggs, salad greens and almond milk in my fridge, so it recommended dishes like egg salad, a grape and almond milk smoothie and a pickle and egg sandwich. Google provided more recommendations and broke them down by the amount of effort level they require. But it really only picked up on the grapes, salad and eggs. 

Aside from answering questions, the Rabbit R1 has a ton of other improvements, such as the ability to answer queries based on previous answers, an improved on-screen user interface when displaying answers, smart timers and alarms and support for additional services like Apple Music and the song generation service Suno. I’ve been able to do things like tell the R1 to set a timer for a power nap (which resulted in a 25-minute timer) and ask it what Spotify recommendations I asked for last week. I was satisfied with its answer; I asked for upbeat indie pop rock that kind of sounds like Olivia Rodrigo, and a Stereogum rock playlist curated for fans of the former Disney star was among its recommendations. 

Where the Rabbit R1 falls short

My biggest concern about the Rabbit R1 is that I simply don’t find myself using it. I’ve been keeping the Rabbit R1 in my purse or backpack regularly for weeks, whether I’m going to a friend’s apartment or the office, but rarely if ever find myself reaching for it. I’ll admit, part of this is out of habit; my phone feels like an extension of myself at this point, so using it for daily tasks is second-nature. 

The larger issue, though, is that phones will likely be able to do everything you can do on the R1 and more — if not today, then soon. At its Snapdragon Summit, mobile chipmaker Qualcomm outlined a vision for how generative AI could enable apps and virtual helpers to execute tasks on your behalf, which sounds a lot like Rabbit’s premise. Google is also prepping a digital assistant that can execute tasks on a person’s computer, according to The Information, which seems a lot like the idea behind Rabbit’s LAM Playground agent (more on that below). And Apple just launched a feature called Visual Intelligence in beta for the iPhone 16, a new mode triggered by the Camera Control button that lets you snap a photo and add it to a prompt. 

Otherwise, the Rabbit R1 has come a long way, but it still isn’t as user friendly as I’d like it to be. LAM Playground, which is arguably the biggest update the R1 has gotten since launch, lets you use Rabbit’s technology to browse the internet for you and complete tasks on your behalf. The company sees its large action model as a step toward building a more sophisticated AI agent. It’s promising but clunky in its current state.

You can use LAM Playground in two ways: through a web browser via Rabbit’s online portal, or on the Rabbit R1 itself. If you’re using it though a web browser, you can type commands and watch as Rabbit executes them in a window on the webpage. If you’re using the R1, you can speak your command the same way you would with any other Rabbit prompt.

Read more: Why Google’s New Android VP Says ‘People Don’t Want to Hear About AI’

One query I tried involved asking Rabbit to go to Reddit and find the top-rated women’s running shoes. That command worked well, but others not so much. For example, when I verbally asked the R1 to go to Google Maps, find New York City bars ideal for a birthday party along with reservation times, the process felt confusing and long. Rabbit kept saying it was checking reservation times, although it never told me which venue it was looking into.

Other times, I was surprisingly impressed by Rabbit’s web browsing skills. I asked it to go to CNET, find the best TV recommendation and see how much it costs at Best Buy. Sure enough, it suggested our top choice, the TCL QM8, and noted that it’s on sale for $1,000 at the electronics retail giant. 

But using LAM Playground isn’t as seamless as a regular voice prompt; the response times are slower, and Rabbit recites each step — which is reassuring, but can also feel awkward. LAM Playground is also inconsistent since it sometimes encounters errors when browsing websites. Yelp, for instance, correctly suspected a human wasn’t navigating the site and blocked Rabbit from finding a result. It’s also worth noting that Rabbit calls LAM Playground an “experiment” on its website, and it very much feels that way at this stage. 

Another one of my gripes is that the R1 is not an ideal interface for actually putting Rabbit’s recommendations into practice. If I want to revisit links, recipes or Spotify recommendations from previous prompts, I have to go to Rabbit’s web portal on my phone or laptop, which certainly weakens Rabbit’s argument for an app-free future. 

Rabbit never claimed that the R1 would replace your phone, but it did insist that the device could do certain things better than your phone — and then failed to deliver on those promises initially. Even with its myriad improvements, all indications point toward a future in which getting things done on our phones is less about opening apps and more about issuing commands. All of this makes the R1 feel like a product that’s somehow ahead of its time, but also not moving fast enough at the same time.