Skip to content
Provider directory
Modal logoModal

Serverless GPU-first compute with memory snapshots.

Run in a trial

Objective

—

speed · reliability · cost

Measured runs

—

untested

Median

—

total wall time

p95

—

tail latency

Success

—

completed runs

Avg cost

—

per run

Runtime

container

gVisor

State & compute

Stateful · Managed runtime

GPU available

Commercial shape

$0 idle

per-second, scale-to-zero

No measured runs yet.

Modal remains visible as untested until real evidence arrives. A simulation never becomes public benchmark data.