Richard is the PC Hardware Lead at XDA and has been covering the technology industry for almost two decades. He's been building PCs since young, and when not creating content, you can often find him inside a chassis somewhere.
Microsoft Copilot hasn't got the best rep. What was designed to be a convenient feature to make AI more accessible for those less tech-savvy, it actually turned into something many (myself included) loathed. What's more is my current laptop, the trusty Lenovo Yoga Slim 7x Gen 9, a Copilot+ PC with a Snapdragon X Elite CPU, is primed for running local models. With PowerToys and a little tinkering, I managed to make this dedicated Copilot button on my keyboard more useful by loading a local LLM.
The problem with Microsoft Copilot
It's not really local is it?
The dedicated button on my laptop's keyboard was enticing to me. No matter which app I was using, I could press this key and fire up a local instance, utilizing the excellent capabilities of the included chip. Boy, was I wrong. All this thing does is launch Microsoft's Copilot app. So far, so good. But this is entirely cloud-based. What's the difference between running ChatGPT, Claude, or some other cloud-based solution? Many of us already pay for one such service.
That's where the Copilot implementation started to fall apart for me. The laptop in question has that mighty Snapdragon X Elite processor, based on Arm and rocking an integrated NPU that can actually perform. This Qualcomm Hexagon NPU is rated at 45 TOPS, which means very little, so I wanted to put it to the test with a locally hosted LLM. The model I have has just 16GB of RAM, so we won't be going wild, but even a small yet useful model can work wonders.
I already made good use of Ollama and LM Studio on other devices, and it was something I toyed with on this very laptop, so I never found myself using Copilot at all. The Deepseek-r1 model is brilliant, notably 8b. The same goes for Qwen3.5 with both 4b and 9b options. For testing, I opted for qwen3.5:4b. It easily fits inside the constraints of the 16GB RAM. I don't tend to have long-winded conversations, so context windows won't be a problem, but it also means I can actually have other apps running on the laptop.
That's the entire point of this experiment. To launch an LLM window with the single keystroke, allowing me to expand horizons while handling other tasks. Having a beefy model that soaks up all the RAM and page file would defeat the purpose and slow things down to a halt.
Why running local AI is important to me
I'm a firm believer we should run LLMs on local hardware, be it the device you're actively using or on a specific device within a home lab. Keeping prompts, replies, and processing all on hardware you own ensures you maintain control over everything. It's similar to why I prefer to store all my files on a NAS. I trust my LAN more than Google, Apple, Microsoft, and other big tech giants. The same goes for LLMs and other agents, and I'd much prefer to rely on open-weight models without remote inference.
Saving money, no additional subscriptions. It sounds like the ideal AI implementation, and having it just one keystroke away should be brilliant.
Changing the Copilot key with PowerToys
Creating my own local AI hotkey
PowerToys is a free tool available for Windows. This handy collection of tools allows for the OS to be customized beyond what the traditional control panel offers. All that was required was for Keyboard Manager to be enabled and then create a new remap for the shortcut tool. This was then linked to LM Studio, so every time the key is pressed, the app opens, and I'm good to go with a fully local version of Qwen. After saving everything, I tested the keystroke, and it worked.
Instead of opening Copilot, the AI key now fired up LM Studio, loading qwen3.5:4b. Interestingly, there's a small delay with Copilot and the same is experienced with LM Studio as the model reserves enough RAM. It's only for the first prompt and is something I grew used to with self-hosting Open WebUI with the same occurrence. The only drawback with PowerToys is the app needs to be running for the remap to be effective. If I don't want it to run in the background, Copilot once again starts up.
Why not just change it within the Settings app? The only choice I see when setting the shortcut key is Copilot. No LM Studio or anything else.
The local AI is useful, but it doesn't beat cloud
Nothing really stands against flagship models from the likes of OpenAI and Anthropic. It's why we're struggling with GPUs, memory, and tech availability and pricing in general. These companies are hoarding all these important components to ensure their cloud-based models have ample resources to handle prompts in record time. Firing up Claude or ChatGPT will see almost instant responses with reasoning, thinking, and external tools used. It's impressive.
The same cannot be said for my local Qwen model running in LM Studio on the Lenovo Snapdragon Copilot+ PC. The 16GB of RAM is a limiting factor. It's also standard laptop memory, which doesn't have the best bandwidth for running AI. A GPU would be better, but even then it still pales in comparison to what's available with cloud LLMs. It all boils down to what's expected from the models and how one interacts with them.
Microsoft should alter the Copilot key
This single shortcut should be configurable. I'm still questioning the motives behind making a dedicated AI key tied to Microsoft's own Copilot, cloud-based AI. I get it; it's Microsoft, and the company wants you to use their offering, but it's not really Microsoft AI. Yes, the company is developing its own LLM, but it's based on OpenAI's GPT. If you've used ChatGPT, Copilot feels very similar. And like ChatGPT, Microsoft offers a freemium model with its Copilot implementation.
AI is increasingly becoming a category of software instead of one simple app. The way we're using AI and interacting with these tools is continuing to evolve, and a move like this isn't necessarily about rejecting Microsoft's Copilot but instead allowing me to decide how I want to go about using AI on my own Windows PC. My Lenovo laptop shipped with a dedicated Copilot key, but pressing it now opens the AI assistant I actually want to use. This is how it should have been in the first place.