Provisioned Throughput
Contracted serving capacity for open models, bought in PTUs of tokens per minute, with OpenAI- and Anthropic-compatible endpoints.
Find alternatives to Provisioned Throughput
What it does
- Model inference
- Model hosting
Sources
- Pricing
-
Pricing page: "Inside your contracted capacity you are billed on PTUs rather than tokens." at "$0.05" per PTU per minute, with "PTU quantity and contract term" arranged with sales.
- Description
-
Official documentation · 16 Nov 2026
Page quote: "Contract the throughput you need in PTUs, and it holds even when the fleet is busy. VESSL Cloud runs the engines and the GPUs.".
- What it does
-
Official documentation · 16 Nov 2026
Page quote: "Contract the throughput you need in PTUs, and it holds even when the fleet is busy. VESSL Cloud runs the engines and the GPUs.".
- Sold within
-
Page fetched 2026-11-16 (curl, HTTP 200), page title: "Provisioned Throughput: guaranteed throughput for open models | VESSL Cloud". The page does not say how it is packaged or which offering it sits inside.
- Name
-
Official documentation · 16 Nov 2026
Page fetched 2026-11-16 (curl, HTTP 200), page title: "Provisioned Throughput: guaranteed throughput for open models | VESSL Cloud". The page names it: "Provisioned Throughput".
- URL
-
Official documentation · 16 Nov 2026
Page fetched 2026-11-16 (curl, HTTP 200), page title: "Provisioned Throughput: guaranteed throughput for open models | VESSL Cloud".
- Deployment
-
Official documentation · 16 Nov 2026
Page quote: "VESSL Cloud runs the engines and the GPUs.".
- Interfaces
-
Official documentation · 16 Nov 2026
Page quote: "Keep your OpenAI or Anthropic client. Only the connection settings change.".
Also from VESSL AI
-
VESSL Cloud Infrastructure service
VESSL Cloud rents NVIDIA Hopper, Ampere and Ada GPUs per GPU-hour as self-serve Container Compute, running Workspaces and Jobs metered by the second; reserved VM Clusters and Provisioned Throughput are separate products.
-
VM Cluster Infrastructure service
Dedicated multi-node GPU clusters on H100 through Blackwell, joined by InfiniBand with root on every node, reserved for a contracted term.
Open VESSL AI in the directory
Something wrong here? Send a correction — quote this product id: vessl-ai-provisioned-throughput.