Sizes
When creating a session you can choose a hardware size with the sizeId field. Larger sizes have more CPU and RAM, which increases cost. The default size is micro.
Available sizes
| sizeId | vCPU | RAM | Small-hours multiplier | Overage rate (per hour) |
|---|---|---|---|---|
micro | 1 shared | 1 GB | 0.53x | $0.032 |
small | 1 shared | 2 GB | 1.0x (reference) | $0.060 |
standard | 2 shared | 4 GB | 2.0x | $0.120 |
large | 4 shared | 8 GB | 4.0x | $0.240 |
max | 8 shared | 16 GB | 8.0x | $0.480 |
The Small-hours multiplier is how many Small-hours the session consumes per wall-clock hour. small is the reference unit at 1.0. A microsession uses 0.53 Small-hours per hour; a max session uses 8.0.
The warm pool covers Python sessions on micro size only. Starter plans get 1 warm slot (Python, Micro size), Growth plans get 10 warm slots (Python, Micro size), and Scale plans get 20 warm slots (Python, Micro size). Free plans have no warm pool. Other languages, non-micro sizes, sessions with persistent volumes, and sessions with broad egress enabled always cold-start (roughly 12 to 22 seconds).
Billing
Usage is metered in Small-hours. One Small-hour equals one hour of small-size session time. Each plan includes a monthly allowance of Small-hours. The overage rates above apply when the allowance is exhausted. See Usage and Pricing for more detail.
Choosing a size
Start with micro. If your workload hits memory limits (OOM kills or out-of-memory errors) or runs noticeably slower than expected, step up to small or standard. Most interactive code-execution and light CI workloads fit comfortably in micro.
dotnet build or dotnet test operations in a single session, use small (2 GB) or larger. The micro size (1 GB) can exhaust memory during intensive build or test cycles.