Open-source AI · Data stays in your country
Your own private AI, deployed in one click
Run the best open-source AI models on GPUs in Asia, for your whole team. Your chats, documents and pictures are stored in the country you choose, and never sent to outside AI companies.
- No credit card to start
- Free on your own GPUs
- OpenAI-compatible API
How it works
Pick a model
Choose from the best open-source models, already tested on real GPUs. Switch on what you need: documents, voice, web search.
Pick a country
Your AI runs, and your data is stored, in the country you choose. Or on your own servers.
Click deploy
About a minute later your team is chatting in a private workspace, and your apps can use the same AI through its API.
What your team can do
A private ChatGPT-style workspace, with everything running on open-source models.
Chat and write
Emails, reports, translations and answers, in Chinese, English, Malay, Thai, Vietnamese and 100+ more languages.
Ask your documents
Attach PDF, Word and Excel files and ask questions. Answers point to where they came from.
Read pictures
Invoices, receipts, screenshots and charts: the AI reads them and pulls out what you need.
Speak and listen
Talk instead of typing, and hear answers read aloud in a natural voice.
Create pictures and video
Posters, product shots and short video clips from a sentence, or edit a photo you upload.
Analyse data and make files
Charts from spreadsheets, and Excel, Word, PowerPoint and PDF files to download.
Your data stays yours
Built for businesses that can't send their data abroad: banks, healthcare, government and anyone under data-protection laws such as Malaysia's and Singapore's PDPA, Indonesia's PDP Law or India's DPDP Act.
✓ Locked, and kept in your country
Your browser locks chats and files with your workspace's key, then they're stored in the country you choose, or on your own server. Stored, we can't read them.
✓ GPUs don't store your data
Our software keeps what you send in the GPU's memory while the AI answers, never on its disk or in logs. The machine does see it while it works, so for confidential work use your own server.
✓ Never used to train AI
Only open-source models run here. Your data is never sent to ChatGPT, Gemini or any AI company, and never used to train any model.
✓ Yours to take or delete
Export everything, or delete it all, any time. Every action in your workspace is in its activity log.
On your own servers
Install our agent on your GPU servers: models, chats and files never leave your building.
In verified data centres
Data centres we have checked, in the countries you allow. For data that must stay in-country; you still trust the operator.
On the open marketplace
GPUs from hosts across Asia at the lowest prices, for non-sensitive work.
AsiaGPU Box: everything in your office
For clinics, law firms and auditors who want nothing to leave the building. One GPU server in your office runs the whole platform, and your data is never used to train AI.
✓ One server, set up in an afternoon
A GPU server in your office, installed with one command. Your staff open it in their browser like any website.
✓ Your data stays in the building
The AI, chats, files and accounts all live on that server, and keep working with the internet unplugged. Data is locked in each person's browser before it's saved.
✓ Ready for your work
Chat, turn recordings into text, read invoices into Excel, summarise contracts, and teach the AI your own documents.
Have GPUs? Earn from them
Data centres, companies and anyone with NVIDIA GPU servers in Asia can rent them out here. You keep 85% of every hour they're used. Your machines only compute: customers' chats and files are never stored on them.
Create an account
Free. Then open My machines and click Add a machine.
Run one command
Paste it on your Linux GPU server. It installs everything needed and connects the machine.
Set a price and earn
Choose the country and a price per GPU-hour, list it, and get paid for every hour it's rented.
Tested open-source models
Every model is run on real GPUs before it's offered, and pinned to the exact version we tested. New, better models are added as they come out.
Alibaba's 8B general model. Strong in Chinese, English and Southeast Asian languages; can think step by step before answering.
Alibaba's 8B vision model: everything Qwen3 8B does for everyday work, and it reads pictures too (invoices, screenshots, charts, photos).
OpenAI's open-weight reasoning model (21B, 3.6B active per token, so it answers fast). Thinks before answering; good at tools and coding. Strongest in English.
Google's 12B model in its 4-bit version (quantised by Google, so quality is kept): reads pictures, 35+ languages. Light on memory: leaves most of a 24 GB card free.
Mistral's 14B model (FP8): everyday chat, tools and pictures. Strong in European languages, also Chinese, Japanese, Korean and Arabic.
Alibaba's 6B picture model: photorealistic images, posters and illustrations from a text description, in a few seconds. Understands Chinese and English prompts, including text in pictures.
Black Forest Labs' small picture model: creates pictures from a description, and edits photos you give it ("make the background white", "add a red scarf").
Alibaba's video model: 5-second 720p clips from a description, or from a picture you give it. Each takes a few minutes.
Simple pricing
Your own GPUs
Free
Connect your servers with one command and deploy on them at no charge. Private mode included.
Our GPUs
Per GPU-hour
Prepaid credit, billed only while a deployment runs. Small models share a GPU, so you pay only for your share.
Host your GPUs
Keep 85%
List your GPUs and earn from every hour they're rented. They only compute: no customer data is stored on them.
Questions
Is my data used to train AI?
No. Your data is only used to answer you. The models are open-source and fixed: nothing you send changes them or is kept by them. Built for clinics, law firms and auditors who must keep client data confidential.
Do I need technical skills?
No. Pick a model and a country, click deploy, and chat in your browser. Developers can also use the same AI from their own apps with an OpenAI-compatible API.
Where exactly is my data?
In the data location your workspace chooses: an encrypted storage server in that country, or your own server. Your chats and files are not kept on the GPU servers or on our control servers. While the AI works on them, the GPU machine (and, for rented GPUs, our gateway on the way) handles them unlocked: see 'Who can see your data' in the docs.
Which AI models can I use?
Open-source models only, such as Alibaba's Qwen. We test each one on real GPUs before offering it, keep only the best current ones, and give at least 60 days' notice before retiring a model.
How much does it cost?
Free on your own GPUs. On ours you pay per GPU-hour from prepaid credit, only while a deployment runs. Several small models can share one GPU, so you pay for the share you use.
Can I use my own GPU servers?
Yes. Run one command on each server to connect it. In private mode your prompts and files don't pass through our infrastructure: we send start and stop commands, and the website your browser loads still comes from us (on an AsiaGPU Box, from your own server).
Bring private AI to your team today
Set up in minutes. Your data stays where you decide.
Create your workspaceAsiaGPU runs open-source models only. Model names belong to their owners.