PortableMind Firmware 1.6: Bring Your Own Model
Firmware 1.6 is out. The headline: PortableMind now ships a full Model Manager inside the desktop app. You can install any model from the Ollama registry onto the drive, delete anything with one click, and see exactly how much space you have left. Before this release, you got the presets we picked and that was the end of it. We did not want to be your gatekeeper for what models you could run on your own USB. This release closes that gap. There is also a new Light preset for very old or very cheap machines. If you have been sitting on a 5 year old laptop with no dedicated GPU, 1.6 is the release that finally makes PortableMind useful on it.
TL;DRFirmware 1.6 adds a full Model Manager: install any Ollama model onto the drive, one-click delete, VRAM-aware warnings, plus a new Light preset that runs on any hardware.
What actually shipped
The Model Manager lives inside the PortableMind desktop app as an overlay. No terminal. No ollama pull commands. No copying files by hand. You open the app, click Model Manager, and everything you can install or remove is right there.
Each model shows up as a card with its size, its VRAM requirement, and either an Install or Delete button. When you install, you get a real streaming progress bar with the download percentage and the megabytes transferred. When you delete, you get a confirm dialog so you cannot wipe a 5GB model with a stray click.
At the bottom of the overlay, a disk usage counter tells you how much of the drive is currently taken up by models. That number moves in real time as you install and delete.
The model cards in 1.6
Five cards ship with the Model Manager by default. Every one of them is honest about what hardware it wants.
- Smart Premium · about 5GB · wants 8GB+ VRAM · the strongest chat model in the lineup
- Smart Default · about 4GB · wants 4GB+ VRAM · the balanced pick most people should start with
- Fast · about 1.8GB · wants 2GB+ VRAM · when you care about tokens per second more than depth
- Light · about 1GB · runs on any hardware · new in 1.6, works on machines with no dedicated GPU
- Scout Vision · about 4.5GB · wants 4GB+ VRAM · the multimodal model powering SCOUT
VRAM warnings that actually mean something
The Model Manager reads your hardware on startup and flags cards that will not fit on your GPU. If Smart Premium wants 8GB and your card has 4GB, the card gets a red border and a plain English note that says the model may not fit.
You can still install it if you want to. Sometimes you know something the app does not, like that you plan to run CPU only. But the default is to warn you before you spend twenty minutes downloading a model that will crash on load.
Bring your own model
Beyond the preset cards, 1.6 lets you type any model tag from the Ollama registry and pull it onto the drive. Same streaming progress bar, same VRAM check, same disk counter. Once it finishes, the model is on your USB and runs like anything else.
After install, the routing layer picks it up automatically. AUTO mode will consider the new model on the next request. No restart, no config file, no reboot.
We are not going to pretend we tested every model on the registry. Some will run great, some will be slow, some will be quirky. But the door is open. That was the point of this release.
Why we opened the drive up
The pitch for PortableMind has always been offline AI you own. A curated preset list was a reasonable place to start, but it was also the one part of the product where we still stood between you and the choice.
1.6 fixes that. The presets are still there for people who want a sane default. The Model Manager is there for people who want to swap in something else. Nobody is locked in, and nobody has to open a terminal to get out.
It is a small change on the surface. Under the hood it is the biggest step the app has taken toward your USB, your rules.
How to update
Open the PortableMind desktop app. If you are on 1.5.x, the updater will prompt you when you launch. Accept the update, restart the app, and the Model Manager button will be in the sidebar.
If you are on an older build, plug in your drive and grab the latest installer from the update page. The drive contents are preserved through the update. Nothing you have on the USB gets touched.
Ready to run AI offline?
PortableMind is the plug-and-run offline AI USB with three tiers: CORE ($49, Windows, chat), v1.7.1 ($79, voice, multi-modal chat, SCOUT, in-app Model Manager), and MAX-SPEED for power users. No internet, no subscription. Pick the tier that fits your needs.
Conclusion
If you have been waiting to try a bigger model, or if you have been stuck on a slow machine that could not handle any of the presets, 1.6 is the release that unblocks you. Update the app, open the Model Manager, and grab whatever fits your card. The drive is yours now.
Read the full 1.6 release notesFrequently asked questions
Long-tail answers for the search queries around this topic.
- Does the Model Manager work offline?
- The manager itself needs internet exactly once, to pull the model from the Ollama registry onto your drive. After that model is on the USB, it runs fully offline forever. Nothing phones home during use.
- What if I never have internet at all?
- The drive still ships with four curated models preloaded out of the box: Smart Default, Fast, Scout Vision, and the new Light preset. You can chat, use voice, and use vision without ever going online. The Model Manager is only for pulling extra models beyond what shipped.
- Can I break the drive by deleting things?
- No. Every delete requires a confirm dialog, and the main pm-* presets can always be reinstalled from the Model Manager with one click. There is no state you can wedge yourself into that a reinstall does not fix.
- Does this cost extra?
- No. Firmware 1.6 is a free update for every existing owner. The Model Manager, the Light preset, the VRAM warnings, all of it. There is no subscription, no tier gate, no pay-to-unlock. If you own the drive, you get it.
- What is the new Light preset for?
- Light is a roughly 1GB, 1.5B parameter model that runs on basically anything, including old laptops with no dedicated GPU and thin machines with 4GB of RAM. It is not going to write your novel, but it will chat, answer questions, and give you a working offline AI on hardware that could not touch the other presets.
- Which models are safe to pull from the Ollama registry?
- The registry is the model source we tested against, and any of the mainstream chat and code models on it should install cleanly. That said, we do not ship guaranteed support for arbitrary GGUFs from anywhere on the internet. Stick to the Ollama registry and you will have a smooth time.
- How much space is on the drive?
- Depends on your tier. The Model Manager shows total disk usage at the bottom of the overlay, so you can see exactly how much room you have left before you start a big download. If a model would push you over, the app tells you before it starts.
- How does AUTO mode know about a model I just installed?
- The routing layer scans the drive on the next request, so a freshly installed model shows up in AUTO mode automatically. You do not need to restart the app, edit a config, or pick the model manually.
Carson Dresser created the first offline AI USB. He builds from South Florida, where hurricane season makes losing power a lived reality rather than a thought experiment, and he watched cloud AI fail people at the worst possible moments: locked accounts, surprise price hikes, assistants that vanish with the signal. In 2025 he decided intelligence should be something you own, and shipped it. Working alone, he engineered the firmware, tuned the local models, wrote the desktop app, and built the over-the-air update system that delivers cryptographically signed payloads to drives already in the field. He also designed the product, built this website, and runs the store, the kind of range that usually takes a company. Before PortableMind he built ClipStitcher and a stack of practical automations for creators and small operators. Every order that ships and every support email still goes through Carson personally.