NVIDIA L4: built for inference and compute
The attached GPU is an NVIDIA L4 with 24 GB of GDDR6 VRAM and up to 242 TFLOPS in FP16 (485 TFLOPS in FP8). Standard NVIDIA drivers are pre-installed: CUDA and PyTorch work out of the box.€0.70 per hour, billed by the minute
The GPU costs €0.70 excl. VAT per hour, billed by the minute, with no commitment. You attach it, you compute, you detach it — billing stops the minute the GPU is released.A GPU for every workload
Train a model, serve an LLM, render a 3D scene, or accelerate a data pipeline: the GPU joins your instance in a few minutes, only when you need it.How it works
- Open your instance in the console and select Attach a GPU.
- Restart the instance so the GPU is taken into account.
- Work — then detach the GPU when you are finished. Billing stops to the minute.