Inference through an API
Connect your application to our model endpoint. Bring your API key and use the models available on the server.
GPU compute & model inference
Good GPUs.
Good company.
You bring the model. We bring the compute.
We’re Shiba Rental, a small team building a home for your next inference workload.
YOUR WORKLOAD IS IN GOOD PAWS.
Small team. Big compute plans.Model inferenceGPU capacityRoom to grow
01 / The offering
A place for
your models to run.
Some workloads need a whole GPU. Some just need a steady stream of tokens. We’re building for both.
Capacity that fits
We’re planning dedicated instances and reserved GPU capacity for predictable workloads, with volume arrangements as we grow.
Tuned for the actual job
Model size, context length, batching, and concurrency all matter. Our approach is to tune the deployment around what you need to run.
02 / Plug in
Meet your
model endpoint.
Use this base URL in your API client. Requests require a valid API key; the models route lists the model IDs your server makes available.
Already have access? Your existing model API key works here.
Base URL
https://shibarental.tail0e03fa.ts.net/v1curl https://shibarental.tail0e03fa.ts.net/v1/models \ -H "Authorization: Bearer YOUR_API_KEY"
Use a returned model ID when sending requests to /v1/chat/completions. Streaming responses are forwarded as they arrive.
03 / Why Shiba Rental?
GPUs shouldn’t
sit around bored.
We’re a new company expanding our own AI infrastructure. When that compute isn’t running our workloads, we want to put it to work for other teams.
That’s the idea behind Shiba Rental: make better use of the machines we operate, and build pricing and capacity around real demand. We’re starting small and choosing the next hardware as the workloads come in.
RTX PRO 6000 · 96GBH100H200GB300-class
Hardware under consideration for our expansion. This is a roadmap, not a list of currently available inventory.
A few useful details
Before you fetch your first token.
Can I rent a dedicated GPU?
Dedicated instances and reserved capacity are part of our expansion plans. Availability will depend on the hardware deployed and the needs of each workload.
Where are the prices?
We haven’t published a rate card yet. Our planned pricing will depend on the model, hardware, volume, and whether capacity is shared or reserved.
What happens to my prompts?
Our customer data commitments are set out in the Privacy Policy. Read it before sending data to the service.