Skip to main content
Common questions about Tensorlake’s pricing model, plan tiers, and the resource limits that apply to sandboxes and image builds. For the live pricing page, see tensorlake.ai/pricing.

​
How is sandbox-as-a-service pricing typically structured?

Sandbox-as-a-service products are usually billed on usage rather than a flat seat fee. The common dimensions are:
  • Compute time — CPU-seconds or vCPU-hours while the sandbox is running.
  • Memory-time — GB-seconds or GB-hours of allocated memory.
  • Storage — disk for snapshots, suspended-state preservation, and persisted volumes.
  • Network egress — outbound bandwidth.
  • Build minutes — time spent building custom images, sometimes metered separately.
Tensorlake Cloud uses per-second sandbox billing with separate CPU and RAM rates, plus per-GB-hour storage for suspended snapshots.

​
Does Tensorlake have a free tier?

Yes. The Free tier is $0 forever with no credit card required. It includes:
  • 2 concurrent sandboxes
  • 1 core / 1 GB RAM / 10 GB disk per sandbox
  • Unmetered sessions
  • Self-serve docs and community Slack support
  • SOC 2 Type 2 compliance
The Free tier’s 1 core / 1 GB RAM / 10 GB disk allowance matches the SDK defaults, so most starter scripts will run unchanged.

​
How much does a Tensorlake Sandbox cost?

Sandbox runtime is billed by the second. One credit is worth $0.01. Snapshots are billed per GB-hour while suspended.

​
What are the Tensorlake plan tiers?

See tensorlake.ai/pricing for current rates.

​
Is sandbox billing per-second or per-hour in Tensorlake?

Per-second. Rates are quoted per hour for clarity (0.05/hrpercoreonOn−Demand,0.05/hr per core on On-Demand, 0.03/hr per core on Pro), but you only pay for the seconds the sandbox is running. This is why suspending a named sandbox between agent turns is meaningfully cheaper than leaving it running idle — see suspend/resume.

​
Do I pay for a suspended Tensorlake Sandbox?

A suspended sandbox consumes no compute, so CPU and RAM time are not billed while suspended. Snapshots are billed per GB-hour while suspended.

​
What are the default resources for a Tensorlake Sandbox?

By default, Sandbox.create() allocates:

​
What are the memory and disk limits for a Tensorlake Sandbox?

  • Memory: between 1024 and 8192 MB per CPU core.
  • Disk: between 10240 and 102400 MiB (10–100 GiB).
Disk size is growth-only: when creating a sandbox from an image, disk_mb can grow the root disk at create time. When restoring from a filesystem snapshot, disk_mb can grow the root disk at restore time. You cannot shrink the disk below the source. See the SDK reference and Lifecycle: resources for the full parameter table.

​
How many concurrent sandboxes can I run on each Tensorlake plan?

​
What is the default sandbox timeout in Tensorlake?

The default timeout_secs is 600 (10 minutes). The maximum allowed value depends on your plan: Verify your free account by adding a credit card — it’s used for identity verification only. Setting timeout_secs=0 requests the plan maximum. When timeout_secs elapses:
  • Named sandboxes auto-suspend — state is preserved and you can resume.
  • Ephemeral sandboxes auto-terminate — the state is gone.

​
What are the build-time defaults for a Tensorlake sandbox image?

Build-time defaults are cpus=2.0, memory=4096 MB, and disk=10 GiB. Override them with --cpus, --memory, and --disk_mb in the CLI, or cpus, memoryMb, and diskMb in SDK options. See Sandbox Images for full image-build parameters.