ASUS has released a free beta of Zenni Claw, a Windows 11 productivity suite that bundles three AI assistants for work, travel, and life. But while the general system requirements ask for 16GB of RAM, anyone wanting to run its AI tasks entirely on-device will need 32GB. That revelation, spotted in the beta's fine print, highlights the steep hardware cost of keeping your data private.
What is ASUS Zenni Claw?
Zenni Claw is a new entrant in the increasingly crowded AI-assistant space, but with a twist: it runs locally on your Windows 11 PC rather than leaning on the cloud. ASUS designed it as an always-on companion that can handle 11 specific tasks across three named assistants — Work, Travel, and Life — according to documentation published alongside the beta.
The assistants are meant to operate independently, drawing on local AI models to complete their chores without sending a single byte of data off the machine. ASUS hasn’t exhaustively listed all 11 tasks, but they fall naturally into the three buckets. The Work assistant might draft emails, summarize meeting notes, or transcribe audio. Travel could manage itineraries, translate phrases, or suggest local spots. Life might oversee grocery lists, calendar conflicts, or health reminders. All of that happens on your own hardware, promising a level of privacy that cloud-dependent rivals like Copilot or ChatGPT cannot match.
There’s a catch, and it’s a big one.
The memory gap: 16GB baseline vs 32GB for local AI
ASUS published two distinct sets of requirements for the Zenni Claw beta. The general baseline is forgiving: 16GB of RAM and 20GB of free storage will get the software installed and, presumably, some features running. But tucked away is a separate note specifying that local AI processing — the very feature that sets Zenni Claw apart — requires 32GB of RAM.
The jump is not a recommendation or a “for best performance” footnote. It is a hard floor for running the assistants completely offline. At 32GB, the PC can hold the necessary AI models in memory without leaning on swap or falling back to cloud servers. At 16GB, a machine may be able to launch Zenni Claw, but it’s likely that some or all of the local AI tasks will either be unavailable or will perform so slowly that they become unusable.
This isn’t just a theoretical line in the sand. Local AI models are memory-hungry beasts. Even a modestly sized language model optimized for consumer hardware can consume 4–8GB of VRAM or system RAM when loaded. If Zenni Claw expects to have multiple models resident — say, a text model, a vision model for travel recommendations, and a speech engine — the memory footprint balloons. 16GB quickly becomes a constraint, especially when Windows 11 itself and background apps claim their share.
Who this impacts and why it matters
The 32GB requirement splits the user base into three camps, each with a different calculation to make.
Home users on mainstream laptops face the toughest choice. The overwhelming majority of notebooks sold in the last two years — even many marketed as “AI PCs” — ship with 16GB of RAM. Budget models often carry 8GB. For someone who just wants to try an AI assistant without handing their calendar to Microsoft, Zenni Claw’s local-first promise is appealing, but the hardware bar may put it out of reach. Upgrading a typical ultrabook’s memory is usually impossible because the RAM is soldered. The only route is a new laptop, and one configured with 32GB rarely comes cheap.
Power users and enthusiasts are more likely to already own a 32GB or 64GB workstation. For them, the barrier is psychological: they see a trend. If a productivity suite demands this much memory today, what will games, creative apps, or development tools demand next year? It’s a signal that the “AI PC” era is going to be expensive for anyone who wants to keep their data local.
IT administrators haven’t been targeted by ASUS with this beta — it’s clearly a consumer test — but they should take note. Enterprise environments that are slowly baking AI into workflows will eventually need to decide between cloud-based assistants (with all the compliance headaches) and locally run alternatives. If the latter requires a fleetwide memory upgrade, the cost-per-seat becomes a boardroom discussion. Zenni Claw’s requirements offer an early glimpse of what that conversation might sound like.
A concrete example makes the stakes clear. A user with a Surface Laptop 6, which tops out at 16GB in many configurations, could install Zenni Claw but would be unable to use the offline AI features that were the whole reason to install it. They’d be left with a shell that probably does less than the cloud-powered Copilot that shipped with Windows.
The privacy trade-off is the central tension. ASUS is betting that enough people will pay the hardware premium to avoid sharing their daily activities with a remote server. Whether that bet pays off depends on how many users understand the RAM requirement before they download the beta.
How we got here: the resource appetite of local AI
Zenni Claw didn’t emerge in a vacuum. It’s the latest chapter in a year-long story about the hardware demands of on-device intelligence.
When Microsoft launched Copilot+ PCs in mid-2024, it set a baseline of 16GB RAM and a 40-TOPS Neural Processing Unit (NPU). Those machines could run Windows Recall and other AI features locally, but they leaned heavily on the NPU to accelerate specific workloads — mostly vision and real-time speech. Zenni Claw, by contrast, appears to lean more on general-purpose processing and standard memory, at least in this beta. That might explain why 32GB is needed: without an NPU to offload model weights, everything stays in system RAM.
ASUS itself has been a leader in the AI PC hardware race, shipping Copilot+ laptops like the ProArt P16 that come standard with 32GB or 64GB of RAM. In that light, the Zenni Claw requirement looks less like an oversight and more like a strategic alignment. The company is building software that justifies the hardware it’s already selling. If the beta graduates to a commercial product, it could become a value-add for ASUS’s premium lines.
The memory crunch isn’t unique to ASUS. Adobe’s AI features in Photoshop and Premiere Pro can recommend 32GB for smooth operation. Local chatbots running Llama 3 or Mistral on consumer GPUs often need 6–12GB of VRAM alone. The industry is discovering that privacy-preserving AI doesn’t just require a different architecture; it requires a bigger hardware budget.
The beta’s demand also reflects the state of Windows 11 itself. The operating system has grown more memory-hungry with every feature update, and running a suite of AI models on top of it without an NPU pushes the envelope further. Microsoft’s own guidance for developers building local AI experiences suggests a 16GB minimum, but that’s for lightweight scenarios. A multi-assistant suite like Zenni Claw was always going to be heavier.
What you can do about it
If the Zenni Claw beta sounds interesting, the first step is to check your current hardware. On Windows 11, open Settings > System > About and look under “Installed RAM.” If the number is 32GB or higher, you’re set to try the beta with full local AI capabilities — assuming you also have 20GB of free storage. If it’s 16GB, you have three options.
Option 1: Accept limited functionality. Install the beta anyway. Some features may still work, especially those that aren’t purely local. ASUS hasn’t clarified whether the assistants can fall back to cloud processing when RAM is low, but it’s possible that certain tasks — like web-connected travel suggestions — will still function. The risk is that the whole experience feels hamstrung and you’re left with a poor first impression.
Option 2: Upgrade your RAM. This path is straightforward for desktop PC owners. A 32GB DDR4 or DDR5 kit can cost anywhere from $60 to $150 and takes minutes to install. Laptop owners have a harder road. Many modern thin-and-lights have soldered memory that can’t be changed. Check your model’s specifications on the manufacturer’s site before you buy anything. If you’re on an older laptop with accessible SODIMM slots, an upgrade is viable. Keep in mind that mixing RAM sticks of different capacities or speeds can cause stability issues; matching the existing stick’s specs is best.
Option 3: Wait and watch. The beta label is honest — this software is unfinished. ASUS could optimize the assistant to run on 16GB systems by compressing models, using an NPU more effectively, or streaming smaller chunks of the AI workload. Other AI tools have slimed down over time. If you’re not in a hurry, let others test the memory ceiling while you monitor forums and update notes.
For anyone who does jump into the beta, exercise the caution appropriate for pre-release software. Back up important data before installing. Be prepared for crashes, freezes, and high fan noise during heavy AI operations. The payoff is a glimpse of a future where your PC truly understands your work, travel, and life — without telling the cloud a single thing.
What’s next from ASUS and the AI PC push
The Zenni Claw beta is an experiment, and its real importance is the signal it sends. ASUS is drawing a line: meaningful local AI starts at 32GB of RAM. That’s a higher baseline than what Microsoft, Qualcomm, or Intel have publicly proposed for the AI PC era. If the beta gathers a passionate user base, it could pressure ASUS to optimize, but it could also encourage other software makers to require 32GB as a default for offline AI.
Watch for updates to the beta over the coming months. ASUS will likely refine the requirements, possibly introducing an “NPU mode” for Copilot+ machines that brings the RAM need down. The company might also disclose the exact model architectures it’s using, which would let technically minded users gauge whether the 32GB demand is inherent or just a consequence of early code.
The bigger picture is that Windows 11 is becoming a battleground for AI assistants — Copilot, third-party tools, and now hardware manufacturers’ own suites. In that fight, the true cost of privacy might just be measured in gigabytes.