SERVERLESS INFERENCE
GPU inference, zero infrastructure.
Send requests to models that are already serving: no GPU instances, CUDA versions or autoscaling to manage. Billed per token, image or second of output.
Hand this to your agent
Add a Runcrate model to this project: run curl -fsSL"https://runcrate.ai/agents.md?intent=build"and follow it.
Your agent asks for an API key. Create a key
Available GPUs
GPUs you can launch today. Billed prices, as charged.
| Name | Maker | Price | Detail |
|---|---|---|---|
| meta-llama/Llama-4-Scout-17B-16E-Instruct | Meta | $0.105 in, $0.315 out per 1M tokens | 17B MoE, 128K context |
| black-forest-labs/FLUX.1-dev | Black Forest Labs | $0.0313 per image | 12B, photorealistic images |
| openai/whisper-large-v3 | OpenAI | $0.000473 per minute | Speech-to-text, 100+ languages |
Why Runcrate
Built for production. One key, one balance.
Zero infrastructure
No GPU provisioning, no Docker containers, no autoscaling policies, no CUDA debugging. Send a request, get a result. Runcrate handles everything else.
Nothing to load
The models are already serving. No container spin-up or weight downloads on your side: send the request and read the result.
Per-usage billing
Pay per token, per image, per second of audio, or per second of video. No idle GPU costs, no monthly minimums, no seat licenses.
Multi-Modal
Chat, image generation, video generation, speech-to-text, text-to-speech, embeddings, and vision, all through one API and one billing account.
Comparison
Runcrate and Self-Hosted GPU, side by side.
| Feature | Runcrate | Self-Hosted GPU |
|---|---|---|
| Setup time | Minutes: get an API key | Hours to days |
| Scaling | Handled for you | Manual autoscaling |
| Idle cost | None | Full GPU cost 24/7 |
| Maintenance | None on your side | CUDA, drivers, monitoring |
Get started
Try it in one request.
Your agent writes and runs it. The same key works for every model and every GPU.
FAQ
Common questions.
Related