SaladCloud

Need to run a batch of transcription or text-to-speech jobs without renting a pricey data center GPU? SaladCloud is a distributed GPU cloud that gives AI and machine learning teams affordable, scalable compute for exactly that kind of work.

Its category sits closest to AI developer tools and AI APIs, and the same capacity can serve LLM inference too.

Pricing is per hour and depends on the card.

Batch priority rates start at $0.064 for an RTX 3060 Ti with 8GB VRAM and climb to $0.294 for an RTX 5090 with 32GB VRAM.

In between you get cards like the RTX 4090 at $0.204 and the RTX 3090 at $0.124.

Every GPU option listed comes with 8GB of memory and 4 vCPUs, so VRAM is the main thing you pick.

SaladCloud also has plain CPU instances from $0.005 an hour, including general purpose, CPU-optimized and memory-optimized types.

Check the batch label before committing, since the quoted GPU rates are for batch workloads.

A freemium model is offered.