An open-source CLI that finds available GPUs across GCP regions and hosts open models in one command.
iwant GitHub repository
1 / 2Overview
What is iwant?
Launch any open weights model on bare metal GPU.
iwant is a free, open-source CLI that finds available GPUs and hosts open-weight models in one command. Run iwant up [MODEL] and it hunts across GCP regions for available capacity, provisions a server, and hands you an OpenAI-compatible endpoint, an API key, and SSH access — all inside your own cloud account. It ships with model recipes for gpt-oss-20b, deepseek-v4-flash, deepseek-v4.1-flash, mimo-v2.6-flash, ling-3.0-flash-fp8, and step-3.7-flash (deployed via vLLM), and also supports Hugging Face repos, gguf files, and LoRAs. Lifecycle commands (iwant auth, iwant list, iwant ssh, iwant down) manage deployments, with a dry-run preview, spot instance option, and an idle auto-shutdown timer (30 minutes by default) to control costs. You bring your own GCP account with billing enabled — the CLI itself is free and GPU usage bills to your account until you run iwant down.
Platforms and languages
iwant Availability
Platforms
Languages
Capabilities
iwant Key Features
One-command deploy
Run `iwant up [MODEL]` to rent a GPU, configure the server, and get an OpenAI-compatible endpoint, API key, and SSH access — all in your own cloud account.
Cross-region GPU hunting
Searches GCP regions for available capacity, e.g. 1xL4 for gpt-oss-20b or 8xH100/H200 for DeepSeek models.
Model recipes
Ready-made configs for gpt-oss-20b, deepseek-v4-flash, mimo-v2.6-flash, and more, deployed via vLLM.
Lifecycle commands
`iwant auth`, `iwant list`, `iwant ssh`, and `iwant down` manage deployments; idle auto-shutdown timer defaults to 30 minutes.
Cost control
Dry-run previews, spot instance option, and auto-shutdown; resources bill to your GCP account only until deleted.
Best for
Who uses iwant?
developers and ML engineers
self-hosting open-weight models on cloud GPUs
Plans and access
iwant Pricing
Free
Free open-source CLI; you pay your own GCP usage.
Common questions
iwant FAQs
What do I need to use iwant?
A GCP account with billing enabled and the gcloud CLI installed. iwant currently supports GCP only.
How much does it cost?
The CLI itself is free and open source. GPU resources bill to your own GCP account until you run `iwant down` to destroy them. An idle auto-shutdown timer (30 minutes by default) helps control costs.
Which models are supported?
Ready-made recipes cover gpt-oss-20b, deepseek-v4-flash, deepseek-v4.1-flash, mimo-v2.6-flash, ling-3.0-flash-fp8, and step-3.7-flash. Custom Hugging Face repos, gguf files, and LoRAs are also supported.
Does Ctrl+C stop my deployment?
No — Ctrl+C only detaches from the session. Your deployment keeps running (and billing) until you run `iwant down`.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
