Open RAM
Open-source AI on rented compute
Run open-source models on rented RAM, GPU and CPU machines, benchmark them on real hardware, and deploy the best setup as an API.
Deploy open-source AI on the right machine, every time.
Rent any machine. Run any model.
Rent CPU, GPU and high-RAM compute by the hour, then run, benchmark and serve open models — Open RAM matches each workload to a machine that actually fits.
What you can do
Everything you need to run open-source AI on rented infrastructure.
Rent RAM, GPU & CPU compute
Browse a live marketplace of machines across providers and regions. Pay by the hour.
Launch cloud AI workstations
Spin up a configured Linux box with Ollama, vLLM, Jupyter and CUDA pre-installed.
Run open models as APIs
Deploy Llama, Qwen, Mistral and more to a private endpoint with one click.
Benchmark on real hardware
Compare quality, latency and cost for any model across CPU and GPU profiles.
One API router for many models
Send a prompt; we pick the right model and machine by cost, speed or quality.
How it works
Rent compute or grab an API key, then use it — in three steps.
1 · Rent a machine or create a key
On the Marketplace, rent a GPU/CPU/RAM machine with SOL — or create a universal API key and top it up in SOL.
2 · Open the Workspace
Your rented machines and your API key both live in the Workspace, ready to use in one place.
3 · Run it
Open a Jupyter notebook (or SSH in) to use a machine, or send prompts to any model with your API key. Stop anytime.
You send a request
A prompt, job, benchmark, or deploy.
Router picks compute
Matches RAM / GPU / CPU to the workload + your strategy.
Runs on a rented machine
A marketplace machine with enough resources.
Open model executes
Llama, Qwen, Mistral, SDXL, Whisper…
Result + cost back
Output, latency, RAM/GPU used, price.
Every action — a prompt, deploy, benchmark or job — is matched to a machine with enough RAM / GPU / CPU to run it.
Pay in $RAM for 50% off — and every token burns.
At checkout you choose SOL or $RAM. Pay in $RAM and it's 50% cheaper. Every $RAM spent goes to the Open RAM treasury and is burned automatically every 10 minutes — so supply only shrinks as the platform is used.
50% cheaper
Pay for compute, API credits and more in $RAM at half the SOL price.
Autonomous burn
Treasury $RAM is burned on-chain every 10 minutes. Deflationary by design — supply only goes down.
Transparent treasury
4unRTJ…Tmk5RM
Payments land in one public wallet, then burn — all verifiable on-chain.
$RAM launches soon — pay in SOL today, $RAM the moment it's live.
Two layers, one platform
A compute layer you rent, and an open-source AI layer that runs on it.
Rent the hardware
Raw infrastructure, billed by the hour.
- Rent CPU, GPU and RAM across providers
- Launch ready-to-use cloud computers
- Run heavy jobs straight from the browser
Run the models
Open models, served and compared.
- Deploy models as private API endpoints
- Benchmark quality, latency and cost
- Route requests through one smart API
When you deploy a model, run a benchmark, use the API router, or submit a job, Open RAM matches it to a machine with enough RAM / GPU / CPU — automatically.
Workload to hardware
How common workloads map to the right kind of machine.
Small chat model
Cheap CPU / RAM machine
Huge Llama / Qwen model
High-RAM or GPU machine
Stable Diffusion
GPU machine (≥16GB VRAM)
Whisper transcription
CPU / GPU machine
Long document prompt
High-RAM machine
Private business workload
Dedicated, high-trust machine
FAQ
Straight answers on renting compute, using it, and paying.
I rented a machine — how do I actually use it?
Can I use the machine from my own computer?
What's the difference between renting a machine and the API keys?
How do I use my API key?
Authorization: Bearer <key>, OpenAI-compatible.How do I pay, and when am I billed?
Are the machines real?
Can I pay with the $RAM token?
What happens to $RAM when I spend it?
Run open-source AI on rented compute.
Pick a machine, deploy a model, benchmark the trade-offs, and serve it through one API — all from your dashboard.