Tiiny AI
Tiiny AI aims to make powerful AI easy to own by bringing large-scale models and agent workflows to personal devices. Its philosophy emphasizes intelligence, privacy and autonomy: users' conversations, data and memories remain under their control.
At a Glance
- Individual AI users
- Developers and researchers
- Small teams
- Privacy-sensitive and regulated organizations
- +3 more
AI Tools by Tiiny AI
(1)
PowerInfer
Local LLM Inference Engine GPU
Discussions
No discussions yet
Be the first to start a discussion about Tiiny AI
Latest News
Kickstarter campaign page reported 2,181 backers and $3,069,202 pledged for Tiiny AI Pocket Lab.
Tiiny Pocket showcased at AI Infra Summit.
Pocket Lab surpassed $1 million in Kickstarter pledges in five hours; Tiiny also announced AgentBox positioning, $1,399 early-bird pricing and a $1,299 website deposit price.
Tiiny AI unveiled Pocket Lab at CES 2026 as a pocket-size, offline AI supercomputer capable of running up to 120B-parameter models.
Products & Services
A pocket-size personal AI computer with 80GB LPDDR5X memory and 1TB PCIe 4.0 SSD, designed to run up to 120B-parameter LLMs locally and offline. It connects to a laptop or PC as a local AI terminal and includes TiinyOS, one-click model and agent deployment, local memory and developer tooling.
The personal AI studio operating environment for Pocket, with model and agent stores, one-click deployment, local workflows and API compatibility.
Open-source high-speed LLM inference software using GPU-CPU hybrid offloading, activation locality, adaptive predictors and neuron-aware sparse operators. The project reports up to 11.69x speedup over llama.cpp in its paper's evaluations.
Tiiny AI's fine-grained/neuron-level sparse-activation technology intended to reduce inference computation while preserving model behavior.
Market Position
Tiiny AI positions Pocket Lab as a personal, private and portable alternative to renting intelligence from cloud APIs: it targets up to 120B local models with no token fees and simplified deployment. Its competitive set includes local-model software such as Ollama and LM Studio, model ecosystems such as Hugging Face, and compact/high-performance AI computers such as NVIDIA DGX Spark and AMD AI MAX systems. Its differentiators are the integrated hardware/software stack, 80GB memory, offline operation, PowerInfer/TurboSparse optimization and agent/API workflow support.
Leadership
Founders
Miles Mi
Founder and CEO of Tiiny AI. Coverage describes Tiiny AI's founding team as engineers with backgrounds from MIT, Stanford, HKUST, SJTU, Intel and Meta.
Executive Team
Miles Mi
Founder and CEO
Founder and CEO of Tiiny AI; associated with the company's personal, private and portable AI-computing vision.
Samar Bhoj
GTM Director
Leads go-to-market messaging and positioning for Tiiny AI's local-first, zero-token-fee AI infrastructure.
Founding Story
The team was formed in 2024 to move capable AI away from exclusive, cloud-dependent data centers and toward personal, private and portable hardware. The initial technical vision was to make large-model inference practical on consumer hardware, beginning with the open-source PowerInfer and TurboSparse projects and later extending that work into a pocket-sized AI computer.
Business Model
Revenue Model
Hardware sales and preorders/crowdfunding for Tiiny AI Pocket Lab, with a local software and model/agent ecosystem. The product is positioned around one-time hardware ownership and zero token fees rather than recurring cloud inference charges.
Pricing Tiers
30% off the US$1,999 MSRP.
20% off the US$1,999 MSRP.
10% off the US$1,999 MSRP.
Locks in a US$1,299 purchase price according to the March 2026 company release.
Target Markets
- Individual AI users
- Developers and researchers
- Small teams
- Privacy-sensitive and regulated organizations
- Healthcare and clinical workflows
- Creative professionals
- Private document analysis and long-context reasoning
- Offline work during travel, remote field work or without internet access
- Always-on personal AI assistants and autonomous agents
- Developer model serving, local AI workflows and AIoT prototyping
- Secure creative work and research assistance
- Small-team local inference without cloud usage fees or a high-end GPU
- Kickstarter backers
- Healthcare/clinical users cited in the official site community examples, including offline clinical knowledge-base and point-of-care workflows