Run 120B.
Own Your AI.

Run 120B.
Own Your AI.

Built to run your agents with zero ongoing fees,
zero friction, and complete privacy.
Built to run your agents with zero ongoing fees, zero friction, and complete privacy.

Tiiny AI Pocket

Tiiny AI Pocket

AI Without the Friction

Tiiny AI Pocket beside a desktop computer

0 Token Fees

Local compute. Unlimited possibilities.

Run your models and agents without metered API costs. Use your AI as often as you need.

TiinyOS: Your Personal AI Studio

From task automation to creative exploration, every AI capability in one clean interface, zero setup required.

Handle complex tasks across your tools.
Put repetitive work on autopilot.
Capture what you watch as organized knowledge.
Bring ideas to life through visuals.
Build working apps from a prompt.
Real Possibilities,
from the Community
Real Possibilities, from the Community
@helldiving_firehawk
Yeah, I kinda wish I'd sprung for a second one when I backed the campaign.... man if I knew then what I know now, the temptation would have been absolutely real
Been using mine a lot at work lately, oddly enough not for coding, but for running Hermes in the background doing component/dev board research for me while I do actual work
...
Works nicely, spend a small amount of money for the planning/reasoning phase of a task and then let the Tiiny act like a token factory doing the grunt work "for free".
@RubyTheJeep
I just thought I wanted to tell you how insanely impressive Tiiny Music is.
Over the weekend I created a fake band, a fake album and even used tiiny to create the album art and then had tiiny music create the tracks. Once I formulated the prompt each track took about 5 minutes to complete.
A few tracks went off the rails but thankfully I could stop it. Great job!
@AntAttack
I consult with technology companies. Many of my clients need me to respect their IP, and I can't work with thier content on public infrastructure.
I've had to carefully divide my work so that much of it can be done on public cloud inference, and specific bits could be assembled and completed using models that would run on my 24Gb Mac Mini M4 Pro. I use OffGrid and a Gemma 4 12B to take notes in meetings, for example.
But this is a game changer for me. I can actually research, draft content and format finshed work product using a 30B class model. And now my meeting notes taker can run on a 30B class model.
@Sdrawkcab
i used openclaw jan-march, hermes march-sept, i just tried using openclaw with the tiiny again... big "wtf?"... i'm now setting hermes up with the tiiny
i moved my daily briefing pipeline over to use tiiny yesterday (news sources -> llm w/persona prompting filtering, summarizing, selection stories -> obsidian vault)...
gotta say, I love the tiiny... low 10s tk/s is enough while its running almost silently... i had an L40s running qwen3.8 at like 35tk/s for the last 2 months and had to wear noise canceling headphones in my own office the entire time now that ive got tiiny i've got my office back, the tk/s difference is 100% worth it because its SILENT and i can BRING IT WITH ME when I travel lol
@DoctorNurse89
Blam. It triggers when I leave, when i get home. It will remind me "since youre leaving home, remember you needed an oil filter #kn256 etc" Things like that
If it doesnt know or understand, it opens a line to tiiny in tailscale and openuiwebportal and tiiny figures it out
It sets up a grammar for correct actions and answers, and done.
I made it a rules and logic based systm, that runs mostly on my Kamrui mini PC, written by tiiny in JS
It tracks, learns, and now im setting up a low latency 100% local project i can load onto an SD card and can run on any pc. Dude... Zero token limits is just.... wow....
@DNSGeek
I loaded qwen-3.8-27b and gave it the GM core rulebook and toolkit documents for the new Discworld RPG, told it to GM a new game.
It did a really amazing job. There were several laughs and the scenes were believable. The die rolls were consequential, but also funny. This model did a fantastic job and we will for sure run this again.
The Expert Take
The Expert Take
The Global Headline
The Global Headline
"LLMs usable with this machine are said to be perfect for "PhD-level reasoning, multi-step analysis, and deep contextual understanding.""
"Operating in the "golden zone" for personal AI, it handles most real-world tasks and scales up to 120B parameters, delivering GPT-4o-level intelligence."
"Tiiny AI has managed to pack in a whopping 80GB of RAM and 1TB of SSD storage into this miniature device in order for it to actually be able to handle intensive AI processing."
"For the first time in AI supercomputing, a pocket-sized device is capable of running up-to a full 120-billion-parameter large language model entirely on-device."
"The device is a mini PC designed to execute advanced inference workloads without cloud access, external servers, or discrete accelerators."
"It is as small as a power bank, offering performance typically associated with much larger and more expensive hardware."
The Journey Ahead
The Journey Ahead
Mar–Apr

Kickstarter Campaign

$ 3,069,202 Kickstarter Pledged
Aug

Kickstarter Shipping Begins

Sep

AI Infra Summit

(Santa Clara, USA)
Nov

SC26

(Chicago, USA)
Jan

CES 2027

(Las Vegas, USA)

Unlock Priority Access
Get a Launch Discount

Tiiny Pocket is launching this November.
Enter your email to get early access and a special discount.

Unlock Priority Access
Get a Launch Discount

Tiiny Pocket

Tiiny Pocket is launching this November.
Enter your email to get early access and a special discount.

FAQ

What is Tiiny Pocket?
Tiiny Pocket is a pocket-sized personal AI device designed to run large open models and AI agents locally, including supported models up to the 120B class.
Connect Tiiny Pocket to your Mac, Windows PC, or other supported device, and Tiiny Pocket handles the AI inference while your computer remains available for your everyday work.
Your models and inference can run locally, with no per-token fees and much greater control over where your data is processed.
Who is Tiiny Pocket for?
Tiiny Pocket is for anyone who wants capable AI running locally — from developers building with models and agents, to professionals working with private data, and users who want an always-on personal AI without ongoing per-token fees.
Does Tiiny Pocket need a computer?
Yes.
Tiiny Pocket is designed to work alongside another device such as a Mac, Windows PC, or supported tablet. It does not use a traditional monitor-and-desktop interface of its own.
You interact with Tiiny Pocket through TiinyOS, supported clients, the SDK, or its OpenAI-compatible API.
Is there a subscription or monthly fee?
No. Tiiny Pocket is a one-time hardware purchase. Running supported models locally on Tiiny Pocket does not require a subscription or per-token fee.
Is Tiiny Pocket easy to get started with?
Yes. Tiiny Pocket is designed to make local AI simple:
1. Connect Tiiny Pocket to your device and power.
2. Choose a supported model from the Tiiny Model Store.
3. Run your model or Agent locally.
No command-line setup is required for normal use. Developers can also connect existing tools through Tiiny's OpenAI-compatible API and SDK.
Does Tiiny Pocket work offline?
Yes. Once a supported model is downloaded from the Tiiny Model Store, it can run completely offline. Local inference, chat, Agents, and supported workflows do not require an internet connection.
Internet access is currently needed for downloading models and certain online features such as Web Search or external Connectors. We are also developing offline TiinyOS update packages for air-gapped environments.
Model import support is also being expanded for users who want to bring in additional models.