Back in April 2023 I was trying to build my first voice agent using the Orpheus/Sesame CSM models. It was exciting until it wasn’t. Between the broken vast.ai instances, confusing UX, and absurd pricing, every platform felt like it was saying “you don’t belong here.”
Out of sheer frustration I started talking to providers, curated a solid network, and built Runcrate. The goal was simple: remove friction from AI compute.
Runcrate works like OpenAI or Anthropic billing with no contracts, no reservations, and no begging support for more GPUs. Whatever is in stock is available to everyone.
Each instance also comes with a built-in “Launch VSCode” button that spins up a browser-based VSCode environment with a custom Jupyter extension ready to go.
I just wanted the GPU cloud I wish I had when I started.
If you’re building AI models or running training workloads, I hope Runcrate helps you move faster and spend less.
Would love your feedback.
Dryft