Run, train & export open AI models on your own computer.
UsingOpen Studio is a desktop app for chatting with, fine-tuning, and exporting the latest open AI models. Your data stays on your computer. One-click install on Mac, Windows, and Linux.
One-time $19 · Pay via Wise · License emailed within 24h
Six things buyers use this for
Not a feature list. These are the jobs people buy UsingOpen Studio to get done — and the reason it pays for itself in the first week.
Chat privately
Ask anything of a local model — your prompts and documents stay on your machine. No data sent to OpenAI, no API keys to manage, no per-token bills.
Train on your docs
Fine-tune on your company's PDFs, code, or notes with one click. Build a model that knows your domain without leaking anything to a third party.
Export to anything
Quantize and export to Ollama, llama.cpp, vLLM, MLX, or GGUF in one click. One workflow, every downstream target.
RAG & documents
Drop in PDFs, Markdown, or whole folders and let the model cite them. Built-in vector store and chunking — no separate database to host.
Images, audio, video
Generate and edit images with SDXL/Flux. Transcribe audio with Whisper. Caption video frames. All locally, all from the same app.
Providers, not lock-in
llama.cpp, vLLM, Ollama, MLX — switch backends without re-learning the UI. Bring your own OpenAI-compatible endpoint if you want to mix local with paid APIs.
Plain-English system requirements
No jargon. If you can run Zoom, you can probably run UsingOpen Studio.
| OS | CPU | GPU | RAM | Status |
|---|---|---|---|---|
macOS Best out-of-the-box experience. Apple Silicon handles larger models and training. | Apple Silicon M1 / M2 / M3 / M4 | Metal (built-in) | 4 GB min · 16 GB+ recommended | Fully supported |
Windows GPU preferred for larger models and fine-tuning. AMD ROCm in beta. | Intel or AMD x64 | Optional — NVIDIA CUDA 6 GB+ recommended | 4 GB min · 16 GB recommended | Fully supported |
Linux Full feature parity. Works headless via CLI too. | x86_64 | NVIDIA CUDA / ROCm / CPU fallback | 4 GB min · 16 GB recommended | Fully supported |
Windows + AMD Inference works today. Fine-tuning on ROCm is beta. | Intel or AMD x64 | AMD ROCm 7000-series | 4 GB min · 16 GB recommended | Partial (beta) |
Falls back to CPU inference automatically. Slower, but you can chat, RAG, and export without a GPU.
Metal acceleration works out of the box — no PyTorch CUDA setup, no Xcode wrangling.
On first launch, UsingOpen detects your GPU, picks the right backend, and selects a model size that fits your VRAM.
Built for the people who'll actually use it
Three groups keep showing up in our inbox. If any of these sound like you, UsingOpen Studio is built for your workflow.
Makers and tinkerers
You want to run the latest open models on your own machine — no API bills, no rate limits, no data leaving your computer.
- Chat with Qwen 3.8 27B, GLM 5.3 Flash, DeepSeek 4.1 Flash and more
- Generate images, transcribe audio, caption video — all locally
- Export any model to Ollama or llama.cpp in one click
- Bring your own OpenAI-compatible endpoint when you want to mix
Researchers and students
You need to fine-tune models on your datasets and ship them to where they'll be used. UsingOpen Studio handles the whole pipeline so you spend time on the research, not the devops.
- Drop in JSONL, CSV, or Markdown datasets — chunking handled
- LoRA, QLoRA, full finetune, and DPO exposed as presets
- Auto-resume from checkpoint if your laptop sleeps mid-run
- Export quantized weights ready for Ollama / vLLM / MLX
Anyone who's tried to set this up before
You've spent a weekend installing PyTorch, fighting CUDA, downloading models from Hugging Face, and writing scripts. UsingOpen Studio is the one-click version of all of that.
- Installer handles Python, CUDA, Metal, and dependencies
- First-launch wizard picks a model that fits your hardware
- Built-in UI for everything — no terminal required
- Updates are free while the project is alive
What's actually inside the app
UsingOpen Studio is six tools in one — wrapped in a desktop UI that a non-developer can use. Click a tab to see what each one does.
chat · glm-5.3-flash · Q4_K_M
Multiple tabs, system prompts, and conversation history — all local. Switch models mid-conversation without losing context. Stream tokens at 50–150/s on a typical laptop GPU.
$19, one time. Yours forever.
No subscriptions, no tier games, no card-required trials. Send $19 USD via Wise, then email us your transaction ID and delivery address, and we'll email your license key + download link within 24 hours.
- UsingOpen Studio desktop app (Mac, Windows, Linux)
- All features: chat, train, export, RAG, image / audio / video
- License key works on any of your machines
- All updates while the project is alive — no renewal, no subscription
- Email support from the team that wrote the code
- Zero telemetry, fully offline after first model download
- Amount
- $19 USD
- Method
- Wise transfer
- Delivery
- License key + download link
emailed within 24 hours
By paying you agree to the Terms.
How to pay — 3 steps, ~2 minutes
- 1Send $19 USD via Wise
Open Wise and send $19 (or the equivalent in your currency) with the UsingOpen payment link below. Wise handles currency conversion at the mid-market rate — no extra fees from us. Your payment carries the reference 085034.
- 2Email us your transaction ID
The Wise payment page has no note field, so email your Wise transaction ID and the address you want your license delivered to, to [email protected].
- 3We verify your payment and email your license
Within 24 hours (usually much faster), we match your transaction ID to the payment, generate your license key, and email it — along with the download link and install instructions — to the address you sent us.
What to include in your email
Send two things to [email protected]:
- Your Wise transaction ID — it's in the payment confirmation Wise emails you.
- The email address you want your license key and download link sent to.
The Wise payment page has no note field, so this email is how we match your payment to your key. Already paid and forgot to email? Just send the transaction ID and we'll sort it out.
We verify each Wise payment manually and email your license key. Usually much faster than 24h — but plan for it.
One $19 payment, the app is yours. Updates are free while the project is alive. No renewal, no auto-charge, no card on file.
Support goes to the team that wrote the app. Email us with anything — install questions, model picks, weird edge cases.
“One-time” — what does that actually mean?
You pay $19 once. You download the app. It keeps working — including all future updates while the project is actively maintained. There is no 12-month clock, no renewal price, no "Pro tier" upsell later. If a year from now we ship a major new version, that update is yours too.
Why Wise instead of Stripe / LemonSqueezy?
Wise keeps the price at $19. Card processing and tax handling add cost on every sale, and those costs would have to come out of the price. Wise is peer-to-peer and works in 160+ countries. The trade-off: no instant key — we deliver manually within 24h.
Latest version: 2026.9.27
Enter the license key from your delivery email and pick your platform. Every download is issued for your key and checked against the SHA-256 shown here. Don't have a key? Pay $19 via Wise — we'll email it within 24 hours.
Enter your license key
- macOS (Apple Silicon)UsingOpen-Desktop-MacOS.dmg75 MB
- Windows (x64)UsingOpen-Desktop-Windows.exe71 MB
- Linux (AppImage)UsingOpen-Desktop-Linux.AppImage229 MB
- Ubuntu / Debian (.deb)UsingOpen-Desktop-Ubuntu.deb75 MB
curl -fsSL https://usingopen.com/install.sh | bashiwr -useb https://usingopen.com/install.ps1 | iexInstall, license, troubleshoot
Short guides for the desktop app — installs, license keys, downloads, and first-launch fixes. The installer and the app handle the rest.
Licensing & downloads
Release notes
Release notes
- v2026.9.27Oct 29, 2026Licensed installers for macOS, Windows and Linux
- Installers for macOS (Apple Silicon), Windows (x64), Linux AppImage and .deb
- License-key downloads with per-download SHA-256 verification
- Windows first-run fix: the update's staged launcher no longer reads as damaged
We answer. Humans wrote this app, humans answer the email.
Email is the one place to get help. Support is run by the team that wrote the code — no outsourced call center.
Email support
Email us with anything — license issues, install help, model picks. Real human reply within 1 business day, from the team that wrote the app.
Questions buyers actually ask
If your question isn't here, email support — we'll answer and add it to this list.
Plain-English legal, on a single page
The TL;DR of every policy a buyer needs to know. Full text lives on the linked sub-pages — but you should be able to decide from this section alone.
Terms of use (EULA)
The EULA governs the $19 license, support, and updates. Pay once via Wise, get a license key + download link by email, use the app on any of your machines. Commercial use included for all buyers.
- One-time $19 USD payment via Wise, single tier, no renewal
- Commercial use included for all buyers
- Updates free while the project is actively maintained
- Branding restrictions: you may not remove UsingOpen attribution
- License key is personal to you — not for resale or redistribution
Privacy
UsingOpen Studio is local-first. We collect what we need to sell and support the app: your Wise transaction details and the email you send us, support emails, and license records. Our site and download endpoints keep standard request logs, including IP addresses, through Cloudflare, Render, and Modal. No behavioural tracking, no marketing cookies, no model or prompt logging.
- Payment: handled by Wise; we receive the transaction details and the email you send us
- License delivery: email containing key + download link, to the address you send us
- License records: issued keys, download counts, and revocations in Firebase
- Hosting logs: standard request logs (IP, user agent) at Cloudflare, Render, and Modal
- Cloudflare Web Analytics: aggregate, cookieless, no PII
- No third-party trackers, no marketing pixels, no fingerprinting
Stop reading. Start chatting with a local model.
Send $19 via Wise, then email us your transaction ID and delivery address. Within 24 hours, we email your license key + download link.