boat vs Modal VM Sandboxes

Both give an agent a real Linux VM with its own kernel, Docker and root. boat saves the disk with no code and can run as long as you choose. Modal VM Sandboxes run up to 24 hours each, carry work forward through filesystem snapshots, and sit inside a large serverless AI platform.

Where boat stands out

A VM that keeps its disk with no code

boat saves the disk every minute while the sandbox runs and once more at each stop. A stopped sandbox has no compute charge, and boat keeps its disk for at least 14 days, longer with a saved snapshot. With auto-stop off on a paid plan (default 1 hour), there is no session cap. A resume is like a reboot, and if a machine fails you lose what was written after the last good snapshot.

Desktop, browser and agents ready to go

Every boat sandbox has a desktop and Chrome at 1920x1080 and 60 fps, with computer tools for agents. Seven coding-agent harnesses are installed, and you prompt, steer and follow them through one API. The optional AI credentials proxy ($1 a day while on, plus $1 per 1,000 requests) keeps the real keys of supported providers outside the sandbox.

More CPU work per second in one public benchmark

In StarSling's run of 3 October 2026 (run 37073680030), the V8 Web Tooling Benchmark at 4 vCPU / 8 GB did 28.06 runs/s on boat, 15.48 on Modal's VM runtime and 12.60 on Modal gVisor. That is about 1.8x the VM runtime. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Providers ran on different hardware in different regions.

The companies and products

boat

boat.dev runs sandboxes for AI agents: each one is a full Linux VM with its own kernel.

You start a machine with one command or one API call, then SSH in, run Docker, open a desktop, or let a built-in agent (Claude Code, Codex, OpenCode and others) work in it. The disk stays between sessions, and a stopped sandbox costs nothing.

Modal VM Sandboxes

Modal is a serverless cloud for AI work, and VM Sandboxes are the version of its sandboxes that run in a virtual machine with their own Linux kernel.

You create a VM Sandbox from Python, TypeScript or Go code with runtime="vm", run commands in it, and terminate it when the job is done. Modal built the runtime on the Rust-based Cloud Hypervisor, so Docker, FUSE, systemd and nested cgroups work, with the same API, images and prices as its gVisor sandboxes. It became generally available on 1 October 2026. It suits teams that want many agent machines next to Modal's inference and training.

Which use cases each one fits

For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “5th cheapest of 365” means 4 of the other 364 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.

Use case and its workloadboat: price rankModal VM Sandboxes: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Both run Docker in a real VM. A Modal VM ends within 24 hours. To keep its files, you snapshot the filesystem and boot a new sandbox from it, or use a Volume. A boat sandbox saves its disk at each stop with no code, and has a built-in desktop to check the app it built.
5th cheapest of 365
$634 a month
27th cheapest of 365
$633 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
An agent that runs all month fits boat with auto-stop off on a paid plan. A Modal VM stops at 24 hours. You set it up: snapshot the filesystem before the end and boot a new sandbox from it each day. Running programs start again in the new sandbox.
6th cheapest of 361
$657 a month
Not ranked: misses a requirement
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
Short code runs fit Modal well: per-second billing on max(request, use) and up to 5,000 containers at once on Team. On boat each start counts against the plan's start cap.
189th of 336, with limits
$2,000 a month
Not ranked: misses a requirement
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
Both fit. A stopped boat sandbox has no compute charge and keeps its disk for at least 14 days, with no code; each start counts against the plan's start cap. On Modal you snapshot each user's filesystem and boot from it when the user comes back; snapshots are kept 30 days by default, or with no expiry.
3rd cheapest of 349
$209 a month
Not ranked: misses a requirement
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
Modal suits huge bursts of short rollouts: Team allows 5,000 containers at once, and Modal publishes no per-start fee. boat caps starts per plan: the $2,000 plan includes 90 a minute, and starts above 270 a minute fail with 429.
39th of 349, with limits
$2,000 a month
163rd of 349, with limits
$11,982 a month
Agent evals
250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month
10th cheapest of 359
$200 a month
Not ranked: misses a requirement
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
Both run Docker builds in a VM. boat exposes /dev/kvm on every sandbox for Android emulators. Modal offers nested virtualization on the Enterprise plan only, as an experimental option. Each CI job on boat counts as a start.
2nd cheapest of 374
$86 a month
170th of 374, with limits
$655 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
19th cheapest of 371
$26 a month
Not ranked: misses a requirement
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
boat ships a desktop and Chrome at 1920x1080 and 60 fps, with computer tools for agents. On Modal you set it up: its runnable example installs Xvfb, x11vnc, noVNC, Chromium and a Browser Use agent.
7th cheapest of 367
$200 a month
Not ranked: misses a requirement
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
7th cheapest of 356
$200 a month
Not ranked: misses a requirement
Scraping & bots (startup scale)
50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month
6th cheapest of 343
$90 a month
Not ranked: misses a requirement
Mass scraping & bots (undetectable)
1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month
73rd of 327, with limits
$1,800 a month
Not ranked: misses a requirement
Parallel red teaming
200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month
3rd cheapest of 351
$200 a month
27th cheapest of 351
$464 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Modal VM Sandboxes have no GPUs; Modal's gVisor sandboxes and functions do. boat has no GPU on any size.
Not ranked: misses a requirementNot ranked: misses a requirement
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Not ranked: misses a requirementNot ranked: misses a requirement

Key features

FeatureboatModal VM Sandboxes
IsolationFull VM with its own kernel, on boat's own AMD Ryzen 7 9700X and Ryzen 9 9950X hosts; shared vCPUsVM with its own Linux kernel on Cloud Hypervisor
Docker, systemd, rootYes, in every sandboxYes; you start dockerd yourself
CPU benchmark (V8 Web Tooling, 4 vCPU / 8 GB, StarSling run of 3 October 2026)28.06 runs/s, rank 2 of 15 (shared), on a Ryzen 9 9950X hostVM runtime 15.48 runs/s, gVisor 12.60, both rank 7 of 15 (shared); measured with cpu=4
List price, 4 vCPU / 8 GB running$0.036 an hour, from a $20 monthly plan (all of it VM time for that month)$0.476 an hour at 2 cores (Modal's unit: 1 core = 2 vCPU) + 8 GiB, billed on max(request, use); Starter is $0 with $30 of credit a month
Start limits$20 plan: 12 a minute, 60 an hour, 200 a day included; extra starts $0.03 to $0.07 each, up to 1,200 a day, then refused; never queued100 containers at once on Starter, 5,000 on Team; no per-start fee and no numeric start rate published
Nested virtualizationYes, /dev/kvm on every sandboxEnterprise plan only, as an experimental option
Longest sessionNo cap once auto-stop is off on a paid plan; auto-stop defaults to 1 hour, up to 30 days24 hours (default 5 minutes); continue in a new sandbox from a filesystem snapshot
Stop and resumeStop saves the disk with no code; resume boots again, like a reboot; disk kept at least 14 days, longer with a saved snapshotNo paused state. You set it up: snapshot_filesystem() before the end, then boot a new sandbox from it (kept 30 days by default, or with no expiry), or keep files in a Volume
ForkFork a running sandbox from its last 1-minute disk snapshot; writes from the last minute or so can be missingSnapshot a running sandbox's filesystem into an image and boot any number of new sandboxes from it
Memory snapshotsNo; a resume is like a rebootYes, in alpha: snapshotting ends the sandbox, snapshots expire after 7 days, no GPUs or region pinning
Disk51 GB on 2 and 4 vCPU sizes, up to 251 GB on xlarge; fixed per sizeFixed 512 GiB root filesystem on VMs
Desktop, browser, computer useBuilt-in desktop and Chrome at 1920x1080 and 60 fps, with computer tools for agentsNot built in. You set it up: Modal's runnable example installs Xvfb, x11vnc, noVNC, Chromium and a Browser Use agent
Coding agentsSeven harnesses installed, with a prompt, steer and events APIYou install the harness in the image; official examples cover the Claude Agent SDK and OpenCode
Keeping API keys out of the sandboxOptional proxy ($1 a day while on, plus $1 per 1,000 requests) for supported model providers and custom HTTPS APIsYou set it up: run a credential-injecting proxy in a Sidecar (experimental), from an official example
Inbound accessUp to 50 HTTPS URLs per sandbox, token-protected by default; raw ports over IPv6 on sandboxes with no IPv4 of their ownEncrypted and raw TCP tunnels, connect tokens, inbound IP allowlist
Outbound rulesOpen internet; no provider-side allowlistBlock all, IP-range allowlist, domain allowlist (beta)
RegionsGermany, Finland and France; you cannot pick one; no US or Asia regionUS, EU and Asia-Pacific plus narrower regions; pinning costs 1.15x to 1.75x
GPUsNoneNone on the VM runtime; GPU sandboxes use gVisor
ComplianceSOC 2 in progress; no HIPAA BAASOC 2 on every plan; HIPAA on Enterprise
Compliance and transparencySOC 2 is in progress. We're open about where we stand: live status and data retention policy.SOC 2
Every feature Battleships tracks (43)
FeatureboatModal VM Sandboxes
Full VM (own kernel)YesYes
Docker insideYesYes
Nested virtualizationYesYes
RootYesYes
SystemdYesYes
Desktop GUIYesWith your own setup
BrowserYesWith your own setup
Computer-use APIYesWith your own setup
SSHYesWith your own setup
Persistent diskYesWith your own setup
Pause/resumeYesNo documented way
Automatic snapshotsYesNo documented way
Fork/cloneYesYes
Live fork (no pause)YesNot published
Memory forkNo documented wayYes
Custom image (Docker or snapshot)YesYes
≥24 h sessionsYesYes
Idle auto-stopNo documented wayYes
Wake on requestNo documented wayNo documented way
HTTPS preview URLsYesYes
Raw TCP inboundYesYes
Public IPv4No documented wayNo documented way
Egress allowlistNo documented wayYes
Secret proxyYesWith your own setup
Secret proxy for any APIYesNo documented way
Preinstalled agentsYesWith your own setup
Agent harness APIYesNo documented way
MCP serverNo documented wayNo documented way
GPUNo documented wayNo documented way
WindowsNo documented wayNo documented way
Arm64No documented wayNo documented way
Self-hostingNo documented wayNo documented way
BYOCNo documented wayNo documented way
Open sourceNo documented wayNo documented way
Scoped API keysYesNot published
Spend limitsYesNot published
Usage per sandbox (API)YesNot published
Audit logsNo documented wayYes
SSONo documented wayYes
SOC 2No documented wayYes
HIPAANo documented wayYes
EU data residencyNo documented wayYes
Zero data retentionYesNot published

Performance

boat28.06 runs/s
Modal VM Sandboxes15.48 runs/s

Both ran in StarSling's run of 3 October 2026 (run 37073680030), on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB. boat did 28.06 runs/s, second of 15. Modal's VM runtime did 15.48 and its gVisor runtime 12.60, both rank 7 (shared). On real repos in the same run, total seconds were 135.8 vs 180.8 (Better-Auth), 980 vs 1,205 (Mastra) and 428 vs 659 (OpenClaw), boat vs Modal VM. StarSling ran Modal with cpu=4, and the sandbox saw 4 CPUs (PR #121). Sized that way, Modal bills 4 cores, $0.76/h, not the 2-core $0.476/h list price used on this page. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Providers ran on different hardware in different regions. ComputeSDK has no Modal VM row and boat is not in ComputeSDK, so we cannot compare start time.

V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Modal VM Sandboxes measured on: VM Sandbox, request = size. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure.

Pricing in five scenarios

Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario. Start limits, per-start fees and plan minimums are in the price, boat's included.

Scenarioboat, a monthModal VM Sandboxes, a month
One coding agent, workdays
4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days
$20 · $20/mo$0 · Starter
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$26 · $20/mo$109 · Starter · with limits
50 agents in parallel
50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside
$317 · $20/mo × 3 seats$1,480 · Starter
Sandboxes for your users
2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once
$200 · $100/mo × 2 seats$579 · Team
1,000 short jobs at once
2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month
$8,800 · $2000/mo · with limits$1,911 · Team · with limits

boat: Plans cost $20, $100, $500 or $2,000 a month, and the whole plan price comes back as VM time for that month. Unused time does not carry over. A 4 vCPU / 8 GB sandbox costs $0.036 an hour, billed per second while it runs, busy or idle. A stopped sandbox has no compute charge, and boat keeps its disk for at least 14 days. Saved snapshots above the included count and the optional AI credentials proxy still charge while sandboxes are stopped. Each plan caps starts: the $20 plan includes 12 a minute, 60 an hour and 200 a day. Create, fork and resume each count as one start. An extra start costs $0.03 to $0.07 from the wallet. A card is required, also for the 7-day trial.

Modal VM Sandboxes: Modal bills VM Sandboxes at the same rates as its gVisor sandboxes: per second, for the larger of the CPU and memory you request and what you use, until the sandbox ends. Starter costs $0 with $30 of credit a month. Team costs $250 a month with $100 of credit. Pinning a region multiplies the price by 1.15x to 1.75x. Egress above the plan allowance (1 TiB on Starter, 10 TiB on Team) costs $0.04 per GiB.

Modal VM Sandboxes: Modal prices CPU per physical core, and one core equals two vCPUs. A 4 vCPU / 8 GiB sandbox is 2 cores + 8 GiB: $0.476 an hour when the request equals the size. The scenario bills on this page use Modal's lowest documented setup: a minimal request, with limits at the workload size, billed on use. That burst capacity needs spare host capacity, so it is not guaranteed. StarSling ran Modal with cpu=4 and the sandbox saw 4 CPUs (PR #121); sized that way, Modal bills 4 cores, $0.76/h.

Limits in those scenarios

A Modal VM Sandbox runs 24 hours at most. Files carry over only through a filesystem snapshot (kept 30 days by default, or with no expiry) or a Volume, which your code sets up. The 24/7 scenario has no break, so Modal needs a snapshot and a new sandbox each day and is marked as not fitting as-is. Modal's scenario bills assume a minimal CPU request that bursts above it; that extra CPU is shared and not guaranteed. boat's limits: auto-stop defaults to 1 hour (up to 30 days, or off on a paid plan). Each plan caps starts. On the $20 plan, 200 starts a day are included, paid extra starts go up to 1,200 a day, and further starts fail with 429. boat never queues a start. For 1,000 short jobs at once, even the $2,000 plan includes only 90 starts a minute, and starts above 270 a minute fail with 429, so boat does not fit that scenario. boat has no US or Asia region and no GPU. A boat resume is like a reboot: memory, running programs and /tmp are not kept. If a boat machine fails, you lose what was written after the last good snapshot.

ScenarioboatModal VM Sandboxes
One coding agent, workdaysNone for this scenarioNone for this scenario
One agent that never stopsNone for this scenario
  • Starter: sessions capped at 24 h
50 agents in parallelNone for this scenarioNone for this scenario
Sandboxes for your usersNone for this scenarioNone for this scenario
1,000 short jobs at once
  • 2,400 starts/day included, paid above that (needs ~6,667)
  • 600 starts/hour included, paid above that (needs ~1,000)
  • starts above 90/min are paid, and starts above 270/min fail with 429
  • shared vCPUs on this tier (Modal (VM Sandboxes) sells dedicated cores separately as “VM Sandbox, request = size”): heavy, sustained CPU use can be slowed by other tenants
boat's own limits
  • Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
  • Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
  • A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
  • Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
  • boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
  • Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
  • Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
  • Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source

Which one should you use?

Use Modal VM Sandboxes when

you run huge bursts of short jobs that finish within a day, need memory snapshots, a large disk, regions outside Europe or strict outbound rules, or already run your models on Modal.

  • Regions and outbound rules. Modal runs in the US, EU and Asia-Pacific, and can block a sandbox's outbound traffic or allow only chosen domains.
  • Large bursts, memory snapshots and a big disk. Useful for huge bursts of short jobs: Team runs 5,000 containers at once, memory snapshots are in alpha, and each VM has 512 GiB of disk.
  • One platform with inference, training and compliance. Useful if you already serve or train models on Modal: sandboxes, GPUs and models sit on one account and one bill.

Use boat when

your agents work for hours or days in a machine they come back to: installs, builds, tests, Docker, emulators and a desktop, on a VM that saves its disk with no code and has no session cap once auto-stop is off.

  • A VM that keeps its disk with no code. A Modal VM runs 24 hours at most and has no paused state. To keep going, your code snapshots the filesystem before the end and boots a new sandbox from it, or keeps files in a Volume.
  • Desktop, browser and agents ready to go. Modal's VM gives you the kernel and runnable examples for the rest: you install VNC, a browser and the agent harness, and to keep keys out of the sandbox you build your own proxy in an experimental Sidecar.
  • More CPU work per second in one public benchmark. Agents spend much of their time on installs, builds and tests. Faster CPU work shortens those waits, and per-second billing charges less for them.

Questions people ask

Is boat cheaper than Modal VM Sandboxes?

Per running hour at a reserved size, yes: $0.036 vs $0.476 for 4 vCPU / 8 GB, from a $20 monthly plan. In this page's scenarios, boat costs less in three of five: $26 vs $109 (24/7 agent), $317 vs $1,480 (50 agents) and $200 vs $579 (sandboxes for your users). For one agent on workdays, Modal costs $0 inside Starter's $30 monthly credit and boat costs $20; to keep files between days on Modal, you snapshot the filesystem and restore it, or use a Volume. For 1,000 short jobs at once, boat does not fit ($8,800 on the $2,000 plan with paid extra starts), and Modal Team costs $1,911 on shared burst CPU. Modal's bills assume a minimal request that bursts, which is not guaranteed.

What is the difference between Modal Sandboxes and Modal VM Sandboxes?

They share one API and one price list. The gVisor runtime can use GPUs. runtime="vm" gives a real Linux kernel for Docker, FUSE and systemd, but no GPUs. If you leave the runtime unset, Modal picks one for you.

How long can a Modal VM Sandbox run?

The default timeout is 5 minutes and the maximum is 24 hours. For longer work, Modal recommends a filesystem snapshot and a new sandbox started from it. boat has no session cap once auto-stop is off on a paid plan (the default is 1 hour).

Can I run Docker and Android emulators in each?

Both run Docker inside the VM. boat exposes /dev/kvm for Android emulators and VMs on every sandbox. Modal offers nested virtualization on the Enterprise plan only, as an experimental option.

Is boat faster than Modal VM Sandboxes?

On CPU work in one run, yes. In StarSling's run of 3 October 2026, boat did 28.06 V8 Web Tooling runs/s, Modal's VM runtime 15.48 and Modal gVisor 12.60, at 4 vCPU / 8 GB. The boat sandboxes in that run were on Ryzen 9 9950X hosts, and boat does not promise a CPU model. Modal ran with cpu=4, which Modal bills as 4 cores. ComputeSDK has no Modal VM row and boat is not in ComputeSDK, so we cannot compare start time.

Is boat SOC 2 compliant?

SOC 2 is in progress. Modal has SOC 2 on every plan. We're open about where we stand: live status at status.ascii.dev and our data retention policy at docs.boat.dev/data-retention.

Start a sandbox

Install the CLI, then run boat new.

7-day free trial

More comparisons