Vast.ai and RunPod connected

The unified interface for GPU processing

We find an available machine, start your workload, and replace it if it fails. You keep one API.

Browse available GPUsSee the API
Capacity liveConnecting
Selected pool

RTX 5090

32 GB
Ready now
234
Starts at
$0.280/hr
Host checks passedUp to 31B modelsVast.ai
Vast.aiRunPodOne inventory
621 machines ready6 GPU types$0.105/hr starting rate inventory updates live
What Gridline handles

You ask for a GPU. We handle the rest.

No provider-specific setup in your application.

  1. 1

    Find a machine

    We check connected providers and filter out hosts that are slow or unavailable.

  2. 2

    Start your workload

    We choose a GPU with enough memory and launch it for you.

  3. 3

    Replace failures

    If the host drops, we move the request to another machine.

Connected nowVast.aiRunPod

Different catalogs, regions, and host types. One list in your application.

See current capacity
Host recovery

A failed GPU should not fail your request.

This stream starts on one machine. When that host drops, another machine finishes the response.

POST /v1/chat/completionsstream: true
RTX 3090 · 7f3a1c · $0.109/hrstreaming
A generation that began on one GPU and was completed by another after the first failed, with no duplicated or missing tokens.
OpenAI-compatible

Change one URL.

Keep your client and model calls. We choose and run the GPU.

client.pyPython
from openai import OpenAI

client = OpenAI(
  base_url="https://api.gridline.dev/v1",
  api_key=GRIDLINE_API_KEY,
)

response = client.chat.completions.create(
  model="gemma-4-12b",
  messages=messages,
)
ReadyPOST /v1/chat/completions
ModelCurrent route
Gemma 4 E4B8,192 context
RTX 3090106 t/s · 44 streams
$0.105/hrGPU cost
Gemma 4 12B8,192 context
RTX 309067 t/s · 15 streams
$0.105/hrGPU cost
Gemma 4 31B8,192 context
RTX 509052 t/s · 8 streams
$0.280/hrGPU cost
Memory checkedHost checkedFallback ready
Get started

Use one API for GPU compute.

Browse live capacity or create an account and send your first request.

Start buildingBrowse GPUs