Vast.ai Serverless
Autoscaling GPU endpoints for AI inference on the Vast.ai fleet: workers are provisioned from load predictions, with a reserve pool for fast cold starts, defined from the Python SDK, CLI or API and billed per second at the platform's GPU rates.
Find alternatives to Vast.ai Serverless
What it does
- Model inference
- GPU cloud
Sources
- Pricing
-
Official documentation · 18 Oct 2026
"Fully transparent with no surcharge for serverless." and "Per-second billing with On-Demand, Interruptible, or Reserved pricing"; live GPU rates are on the pricing page.
- Description
-
Official documentation · 18 Oct 2026
"Automate the provisioning of GPU workers to match the dynamic computational needs of your workloads."
- What it does
-
Official documentation · 18 Oct 2026
"Automatically scale your AI inference up or down based on customizable performance metrics."
- Sold within
-
Read the Serverless page and the pricing page: Serverless runs on the same GPU fleet and is billed at its rates ("no surcharge for serverless"), but no page states whether it is included in or bought apart from the GPU Cloud.
- Name
-
Official documentation · 18 Oct 2026
Products menu entry "Serverless"; page title "Serverless | Vast.ai".
- URL
-
Official documentation · 18 Oct 2026
The product's own page, titled "Lowest Cost, Autoscaling GPU Cloud | Serverless | Vast.ai" and linked as "Serverless" in the Products menu.
- Deployment
-
Official documentation · 18 Oct 2026
Page: "Run workloads on dedicated infrastructure with full environment control and SOC 2 Type I compliance."
- Interfaces
-
Official documentation · 18 Oct 2026
"Define your Docker image, packages, and autoscaling config entirely in code. The Vast SDK handles endpoint creation and management"; "Prefer code? Hit our lightweight CLI or API endpoints".
- Intended for
-
Official documentation · 18 Oct 2026
"Deploy from Python, Not the Dashboard": for developers running inference workloads.
Also from Vast.ai
-
Vast.ai GPU Cloud Infrastructure service
Marketplace for renting GPUs (RTX 3060 to B200 across 68+ types) with on-demand, interruptible and reserved terms billed per second, driven from a web console, API, CLI or Python SDK.
-
Vast.ai Clusters Infrastructure service
Dedicated GPU and CPU clusters for frequent AI, ML and HPC training runs, scaled from tens to hundreds of GPUs, with custom networking and storage, utilisation and cost dashboards, enterprise SLAs and specialist support, arranged through sales.
Something wrong here? Send a correction — quote this product id: vast-ai-serverless.