Virtual machines that boot in 70 milliseconds.
Any Linux distro today. macOS and Windows in closed beta.
Boot time
70 ms
p50, c1m2, cold
Snapshot pause
0 ms
VM keeps running
Branch restore
< 100 ms
per branch
Compute saved
50–90 %
vs equivalent
import asyncio from fastvm import FastVM async def main(): async with FastVM() as client: # launch, 70 ms p50 vm = await client.launch(machine="c1m2") out = await client.run(vm, "python3 --version") print(out.stdout) # => Python 3.13.5 asyncio.run(main())
Benchmarks
Measured against the alternatives.
Boot time, to scale
FastVM≈ 14× faster than Modal70 ms
E2B300 ms
Daytona300 ms
Modal1 000 ms
p50 cold boot, c1m2, May 2026.
Full comparison
FastVME2BModalDaytona
Boot70 ms300 ms1 000 ms300 ms
Snapshot0 ms, zero-downtimeInterrupts procsVM shutdownFilesystem only
StorageSub-linearLinearLinearLinear
Compute50–90% savedStandardStandardStandard
IsolationDedicated kernelDedicated kernelShared, gVisorDocker
Methodology: p50 across 1,000 cold launches against c1m2 instances in sfo, May 2026. Competitor numbers from public docs and our reproductions.
Built for
What people build on FastVM.
Coding agents
Snapshot every step. Roll back instantly on red. Spin up ten parallel attempts at a test suite, let an agent pick the best result, merge it.
Get started
Sixty seconds.
1$ pip install fastvm
2$ export FASTVM_API_KEY=…
3$ python -m fastvm launch
→ vm-7c4d9a ready (72 ms)