From $3 Million Crowdfunding to 10,000 Units Shipped: Yunqi Capital Leads $10M+ Angel Round in "Tiiny AI" | Yunqi Partners

Tiiny AI, a Yunqi Capital seed portfolio company building a Personal Agent OS, recently announced the completion of a multi-million-dollar angel round series, with Yunqi Capital increasing its stake in this round.

Founded just over a year ago, Tiiny has already achieved consecutive leaps from technical conviction to product validation to commercial deployment: its first product, the Tiiny AI Pocket, raised over $3 million on Kickstarter and has now entered mass production and delivery; simultaneously, it is advancing deep OS-level partnerships with tier-one international PC manufacturers, with projected install demand exceeding tens of thousands of units. Learn more with Yunqi Capital.

The following content is adapted from "Z Potentials."

More notable than the funding itself, Tiiny is joining forces with a leading international PC manufacturer for deep commercial collaboration at the OS level. Reportedly, the per-unit license price for TiinyOS in this project exceeds that of a household-name PC operating system, with the first cooperative products scheduled to ship in Q4 2026.

This means that after achieving $3 million in crowdfunding with a 300-gram Tiiny AI Pocket, this startup — barely a year old — is now entering the global PC supply chain, becoming an independent system layer connecting models, agents, data, and devices.

How did Tiiny rapidly reach these critical milestones from product validation to scaled commercial deployment, and why has it attracted concentrated bets from multiple top-tier dollar funds?

Ahead of Consensus, Tiiny Pioneered the Personal Agent OS

From its founding in Q1 2025, Tiiny judged that personal AI would not remain confined to cloud-based chatbots, but would evolve toward locally resident, continuously user-aware systems with ownership vested in the user — a Personal Agent System. While industry attention remained fixated on foundation model capabilities and cloud applications, Tiiny had already begun systematically building for the long-term needs of personal AI: local private token supply, on-device inference and multi-model orchestration, Agent Runtime, local long-term memory, and integrated hardware-software experiences.

Over the past year, the industry has accelerated along the trajectory Tiiny originally foresaw. In June 2026, Apple proposed "Run local agentic AI on Mac"; around the same time, NVIDIA released the RTX Spark, offering up to 1 petaflop of compute and 128GB of unified memory, and partnered with Microsoft to build a next-generation Windows PC platform for Personal Agents. The open-source model and local inference ecosystem has exploded in parallel: Ollama's monthly downloads grew from roughly 100,000 in Q1 2023 to 52 million in Q1 2026, a roughly 520-fold increase over three years; GGUF local model repositories on Hugging Face reached 190,000. ATOM data shows that as of March 2026, global open-source large model cumulative download requests hit 2.04 billion, with sub-50B parameter models accounting for over 90%; sub-10B models have already approached or surpassed the performance of trillion-parameter models from 18 months ago on certain vertical tasks.

All these trends reinforce the industry shift of personal AI from cloud to local, from service invocation to system ownership. Meanwhile, the cloud token costs of continuously running agents have become a real barrier, with persistent massive token consumption from long context, multi-turn planning, memory read/write, and background tasks. Taking Moonshot AI's K3 as an example, at its official API pricing, a heavy personal user generating 8 million input tokens and 3.2 million output tokens daily would incur approximately $65 per day, with monthly costs approaching $2,000.

The convergence of three curves — rising on-device compute, leaping lightweight model capabilities, and climbing cloud token costs — is precisely what Tiiny identified from day one. Around this trend, TiinyOS, centered on local device compute, unifies models, compute, data, and agent workflows into a truly user-owned intelligent system that continuously accumulates and extends across devices.

Delivering Results: From $3 Million Crowdfunding to Tens-of-Thousands-Unit Install Plans with a Global PC Giant

In Q1 2026, Tiiny launched the first embodiment of TiinyOS — the Tiiny AI Pocket. Weighing approximately 300 grams and roughly the size of a power bank, the Tiiny AI Pocket was certified by Guinness World Records as the world's smallest personal AI supercomputer capable of running 120B-parameter large models locally. It completed over $3 million in Kickstarter crowdfunding, breaking $500,000 in 4 minutes and $1 million in 5 hours. Recently, Tiiny AI Pocket has completed pre-production validation and preparation, with Kickstarter orders set to begin delivery by late Q3.

Beyond stable progress toward mass production, Tiiny is engaged in deep OS-level commercial partnerships with multiple industry giants. Specifically, a tier-one international PC manufacturer has already licensed the TiinyOS system; reportedly, the per-unit license price for TiinyOS in this project exceeds that of a household-name PC operating system, with projected install demand for TiinyOS cooperation exceeding tens of thousands of units. Additionally, several major model providers and unicorn teams are in discussions with Tiiny about joint adaptation and commercial deployment of their models with Tiiny AI Pocket products.

From Pocket completing product validation with global users, to TiinyOS entering a top-tier PC manufacturer's tens-of-thousands-unit install plan, the Tiiny team is leveraging exceptionally prescient strategic judgment and formidable execution to pioneer the scaled commercial闭环 for Personal AI.

Full-Stack Breakthrough: A Top-Tier Rare Team Spanning OS, Models, and Chips, Running 100B-Parameter Models at ~10W

TiinyOS has constructed a complete system architecture from local tokens to agent ecosystems. Tiiny comprises a world-class operating systems team — a rare full-stack group simultaneously possessing expertise in operating systems, large language model pre-training, and chip architecture — having performed system-level optimization across model operators, inference frameworks, memory scheduling, SSD access, and底层 chip adaptation. It is precisely this full-stack capability spanning OS, models, and chips that enables TiinyOS to stably and smoothly run 100B-parameter large models with only ~10W of power, without relying on expensive hardware stacking.

At the Runtime layer, TiinyOS handles model and agent hosting, multi-model and multi-agent orchestration, background tasks, permission and state management, tool invocation, and personal long-term memory — allowing agents to run continuously locally, continuously update context, and proactively advance tasks even when the user is offline. Even if users switch models or devices, their accumulated data, memory, preferences, and agents remain continuously owned by them. At the ecosystem layer, TiinyOS will further host a Model Store, Agent Store, and Skill Hub. Users can discover, install, and manage different models, agents, skills, and workflows much like installing software; developers, meanwhile, need not rebuild inference, memory, permission, and tool systems for every chip and device, but can develop, deploy, and distribute on a unified Runtime.

Tiiny's "full-stack" extends beyond technical depth across OS, models, and chips to encompass prescient judgment of industry trends, organizational execution in translating conviction into product, and commercialization capability in driving technology into mainstream supply chains. The closed loop of technology, product, and commerce enables Tiiny to continuously transform non-consensus into verifiable, replicable scaled outcomes — constituting an even rarer team value, and perhaps one reason multiple top-tier dollar funds chose to invest.

Following this round, Tiiny will accelerate global delivery of Tiiny AI Pocket and multi-device deployment of TiinyOS, continue investing in global markets and developer ecosystems, steadily improve Agent Runtime and personal memory capabilities, and drive co-evolution of system and model capabilities.

For more on Tiiny's product capabilities, use cases, and real-world performance, visit the Tiiny AI website or watch overseas tech KOL hands-on experience and deep-dive demo videos:

https://www.youtube.com/watch?v=RkzCAaIV_cQ&t=4s