Starter — 1× NVIDIA RTX (16 GB)

Description

What you get

• 1× NVIDIA RTX GPU, 16 GB VRAM — dedicated passthrough, not time-sliced or partitioned
• 8 vCPUs · 32 GB RAM
• 500 GB local NVMe — fast enough that dataloading does not starve the GPU
• Root access — install PyTorch, TensorFlow, Jupyter, Docker, whatever you need
• Private, isolated environment
• Email and community support
• Deployed within 24 hours of confirmation

Who this is for

Solo developers and small projects. It comfortably handles QLoRA on 7B-class models, LoRA up to about 3B, diffusion fine-tuning at ordinary resolutions, and serving quantised models. It is also the right place to do the unglamorous work — data pipelines, tokenisation, evaluation harnesses — without paying A100 rates for it.

Not sure whether 16 GB is enough? We wrote up how to decide between this and the A100, including when the cheaper card is genuinely the right answer.

Billing

$149 per month in USD, exclusive of tax. No setup fee, no lock-in. Your first month is covered by a 14-day full refund. Instances run on our own hardware in Mumbai, India. Full terms in our Terms of Service.

How to order

We provision each instance by hand so we can confirm capacity before taking any payment. Send an enquiry and we will reply within one business day with availability and next steps.