Paperspace vs Replicate
Paperspace, part of DigitalOcean, rents whole Linux machines with or without GPUs at hourly rates, and you work inside them over SSH. Replicate runs AI models as containers behind an API, on GPUs that scale up with requests and down to zero.
The companies and products
Paperspace
Paperspace is a GPU cloud for machine learning, owned by DigitalOcean since 2023.
Paperspace Machines are full Linux VMs with a persistent disk, a public IP and optional Nvidia GPUs from M4000 up to H100, billed by the hour. Next to Machines it sells Gradient Notebooks, Deployments and Workflows for training and serving models. It is built for people who train, test and run models, not as a sandbox API: its API manages machines, and you work inside them over SSH.
- Isolation
- Full Linux virtual machines
- Operating systems
- Linux; Windows templates closed to new users since 1 July 2024
Replicate
Replicate runs AI models with an API: tens of thousands of community and official models, plus your own models packaged with Cog.
You call a public model with one API request, fine-tune one on your data, or package your own model with Cog. Cog is Replicate's open-source tool that builds a Docker image with an HTTP prediction server. Replicate runs it on CPU or NVIDIA GPU hardware as deployments that scale up with traffic and down to zero. It is built for teams that serve models, not for agents that need a machine of their own, so this page compares the places where the two overlap.
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. “22nd cheapest of 359” means 21 of the other 358 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 workload | Paperspace: price rank | Replicate: price rank | boat: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Paperspace gives a coding agent a real VM with Docker, but its API cannot run commands inside it. Replicate runs a model container reached through its prediction API, and files that must last go to storage you set up. boat ranks 8th in our default preset with a command API and agents built in. | 164th of 365, with limits $1,169 a month | 283rd of 365, with limits $4,340 a month | 5th cheapest of 365 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives on one machine all month keeps its disk on Paperspace and boat. A Replicate instance starts fresh from its image, and each prediction stops at 30 minutes unless support raises it. | 229th of 361, with limits $1,560 a month | 340th of 361, with limits $11,300 a month | 6th cheapest of 361 $657 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short runs suit Replicate's model: work starts on a request, bills per second and scales back to zero. Paperspace sets hourly rates and does not say whether a part hour rounds up. boat's start limits make big bursts a compromise. | 71st of 336, with limits $61 a month | 255th of 336, with limits $295 a month | 189th of 336, with limits $2,000 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Paperspace needs your own SSH layer to run code inside machines, and Replicate gives your users a model endpoint. boat ranks 2nd in our default preset and gives each user a full VM, within its plan's start limits. | 63rd of 349, with limits $428 a month | 310th of 349, with limits $3,705 a month | 3rd cheapest of 349 $209 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 41st of 349, with limits $2,057 a month | 180th of 349, with limits $17,900 a month | 39th of 349, with limits $2,000 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 22nd cheapest of 359 $331 a month | 79th cheapest of 359 $1,440 a month | 10th cheapest of 359 $200 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month A Paperspace VM can host a CI runner with Docker, and Paperspace ranks 7th in our default CI preset; boat ranks 2nd, with Docker and /dev/kvm on every sandbox. Replicate does not document Docker inside a model container. | 12th cheapest of 374 $205 a month | 313th of 374, with limits $864 a month | 2nd cheapest of 374 $86 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Paperspace fits a dev box that never stops, with SSH and a disk that is kept. Replicate documents no SSH and no persistent disk; you reach a model through its API. | 50th cheapest of 371 $63 a month | 312th of 371, with limits $226 a month | 19th cheapest of 371 $26 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Paperspace streams a desktop only on Windows machines new users cannot create; on Linux you install one yourself. On Replicate you can install a browser in your Cog image and drive it inside a prediction. boat streams a 60 fps desktop with Chrome. | 171st of 367, with limits $298 a month | 292nd of 367, with limits $1,296 a month | 7th cheapest of 367 $200 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month | 246th of 356, with limits $149 a month | 332nd of 356, with limits $1,296 a month | 7th cheapest of 356 $200 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 246th of 343, with limits $91 a month | 307th of 343, with limits $439 a month | 6th cheapest of 343 $90 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 75th of 327, with limits $1,822 a month | 221st of 327, with limits $8,775 a month | 73rd of 327, with limits $1,800 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 191st of 351, with limits $102 a month | 309th of 351, with limits $855 a month | 3rd cheapest of 351 $200 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month Both fit GPU eval jobs: in our default preset Replicate ranks 19th at $1,338 a month and Paperspace 20th at $1,454. boat has no GPU on any size. | 23rd cheapest of 102 $1,454 a month | 22nd cheapest of 102 $1,338 a month | Not 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 requirement | Not ranked: misses a requirement | Not ranked: misses a requirement |
Key features
| Feature | Paperspace | Replicate | boat |
|---|---|---|---|
| Kind of product | GPU and CPU VM cloud for machine learning; each machine is a full VM with its own kernel | Model inference API; each model runs as a Cog container | Sandbox cloud for agents: full VMs with their own kernel, on AMD Ryzen hosts dedicated to boat |
| Shell and network access | SSH and a public IPv4 on every Linux machine, any port, with a static option; no managed HTTPS URLs | No SSH documented; one prediction API per model behind API token auth, no own IP or ports documented | SSH, root and systemd; up to 50 HTTPS URLs per sandbox, token-gated by default; public IPv6; the IP changes at each resume |
| Docker and nested virtualization | Full Linux VM; ML-in-a-Box ships NVIDIA Docker, and you can install Docker on others. Nested virtualization is not documented | The model is the container; Docker inside and /dev/kvm are not documented | Docker, root, systemd and /dev/kvm on every sandbox |
| Working inside from code | The API and CLI manage machines but cannot connect to them; you run commands over SSH from your own code | You call the model through the HTTP API or client libraries | API, SDK and CLI to run commands, move files and fork |
| Files after a stop | Disk kept while the machine exists and billed while off; scheduled snapshots at most hourly, up to 10 kept | No persistent disk documented; instances start fresh from the image, so files that must last go to storage you set up. API inputs and outputs are deleted after an hour by default | Disk saved every minute and at each stop; no compute charge while stopped; disk kept at least 14 days, longer as a saved snapshot |
| Copying a machine | Custom templates; the machine must be turned off first | No disk snapshot or fork documented; a deployment adds instances of the same image | Fork a running sandbox in one call, from its last 1-minute snapshot |
| Run time | Auto-shutdown from one hour to one week, and you can remove it | Instances can stay on, but each prediction times out after 30 minutes unless support raises it | Auto-stop defaults to 1 hour; set up to 30 days, or turn it off on a paid plan for no session cap |
| Scale to zero and wake | Machines can shut down after a set time; you start them again yourself | Deployments scale to zero and start again on the next prediction | No wake on request; you resume a stopped sandbox by API |
| Billing step | Hourly rates while on; a finer step is not published. Disk and public IP keep billing when off | Per second by hardware, from prepaid credit that expires after a year | Per second while running, from a monthly plan ($20 minimum) returned as VM time for that month |
| GPUs | Nvidia M4000 up to H100, up to 8 per machine; H100 needs a request | T4, L40S, A100 80GB, H100 and H200, up to 8 per instance; H200 and most multi-GPU need a committed spend contract | No GPU on any size |
| Ready-made models | ML templates such as ML-in-a-Box | Tens of thousands of community and official models for images, video, speech and text, plus fine-tuning | None; your agent calls the model APIs you choose |
| Desktop and computer use | Built-in streaming only on Windows machines, closed to new users; on Linux you install a desktop and VNC yourself | No desktop product; you can install a browser in your Cog image and drive it inside a prediction | Built-in 1920x1080, 60 fps desktop and Chrome stream with computer-use tools |
| Keeping keys out of the machine | No credentials proxy; Gradient secrets are injected as environment variables | Secrets are sent to the model as inputs and redacted from logs afterwards | Optional AI credentials proxy ($1 a day plus $1 per 1,000 requests); with it on, real keys of supported providers stay out of the sandbox |
| Regions | New York, Amsterdam and Canada; you pick one | No region choice documented; infrastructure subprocessors listed in the United States | Germany, Finland and France; no US or Asia region, and you cannot pick one. Snapshot copies can be stored outside the EU |
Every feature Battleships tracks (43)
| Feature | Paperspace | Replicate | boat |
|---|---|---|---|
| Full VM (own kernel) | Yes | No documented way | Yes |
| Docker inside | Yes | No documented way | Yes |
| Nested virtualization | Not published | No documented way | Yes |
| Root | Not published | Not published | Yes |
| Systemd | Not published | Not published | Yes |
| Desktop GUI | No documented way | No documented way | Yes |
| Browser | No documented way | No documented way | Yes |
| Computer-use API | No documented way | No documented way | Yes |
| SSH | Yes | No documented way | Yes |
| Persistent disk | Yes | No documented way | Yes |
| Pause/resume | Yes | No documented way | Yes |
| Automatic snapshots | No documented way | No documented way | Yes |
| Fork/clone | No documented way | No documented way | Yes |
| Live fork (no pause) | No documented way | No documented way | Yes |
| Memory fork | No documented way | No documented way | No documented way |
| Custom image (Docker or snapshot) | Yes | Yes | Yes |
| ≥24 h sessions | Yes | Yes | Yes |
| Idle auto-stop | Yes | Yes | No documented way |
| Wake on request | No documented way | Yes | No documented way |
| HTTPS preview URLs | No documented way | Yes | Yes |
| Raw TCP inbound | Not published | Not published | Yes |
| Public IPv4 | Yes | No documented way | No documented way |
| Egress allowlist | No documented way | No documented way | No documented way |
| Secret proxy | No documented way | No documented way | Yes |
| Secret proxy for any API | No documented way | No documented way | Yes |
| Preinstalled agents | No documented way | No documented way | Yes |
| Agent harness API | No documented way | No documented way | Yes |
| MCP server | No documented way | Yes | No documented way |
| GPU | Yes | Yes | No documented way |
| Windows | No documented way | No documented way | No documented way |
| Arm64 | No documented way | No documented way | No documented way |
| Self-hosting | No documented way | Yes | No documented way |
| BYOC | No documented way | No documented way | No documented way |
| Open source | No documented way | Yes | No documented way |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | Yes | Yes | Yes |
| Usage per sandbox (API) | No documented way | No documented way | Yes |
| Audit logs | Not published | Yes | No documented way |
| SSO | Yes | No documented way | No documented way |
| SOC 2 | Yes | Not published | No documented way |
| HIPAA | Yes | Not published | No documented way |
| EU data residency | Yes | No documented way | No documented way |
| Zero data retention | Not published | Not published | Yes |
Performance
Paperspace and Replicate are not in the public benchmarks we use, so we cannot say which of the three is faster at CPU work. In StarSling's run of 3 October 2026 (run 37073680030), boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15. Every boat sandbox in that run ran on an AMD Ryzen 9 9950X host; boat also places sandboxes on Ryzen 7 9700X hosts and does not promise a CPU model. Both Paperspace and Replicate are chosen mostly for GPU work, which this test does not cover.
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. 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. Paperspace and Replicate are not in the public benchmarks we use, so we cannot say which is faster.
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.
| Scenario | Paperspace, a month | Replicate, a month | boat, 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 | $18 · Core usage | $57 · Usage · with limits | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $63 · Core usage | $237 · Usage · with limits | $26 · $20/mo |
| 50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside | $734 · Core usage | $2,851 · Usage · with limits | $317 · $20/mo × 3 seats |
| Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once | $186 · Core usage | $1,620 · Usage | $200 · $100/mo × 2 seats |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $411 · Core usage | $3,580 · Usage | $8,800 · $2000/mo · with limits |
Paperspace: Paperspace sets an hourly rate per size while a machine is on: C5 (4 vCPU / 8 GB) is $0.08 an hour. It does not publish whether a part hour rounds up, so our short-job figures count run time only. The boot disk (from $5 a month for 50 GB) and the public IP keep billing until you delete the machine, also when it is off.
Replicate: Replicate bills per second by hardware tier, from prepaid credit that expires after a year. The cpu tier (4 CPU / 8 GB) costs $0.36 an hour. Community public models bill only for active time; private models and deployments also bill setup and idle time, and failed or canceled runs.
Replicate: Plans cost $20, $100, $500 or $2,000 a month, and the whole plan price comes back as VM time for that month; unused included time expires. A default sandbox (4 vCPU / 8 GB) costs $0.036 an hour, billed per second. A stopped sandbox has no compute charge, but saved snapshots above the included count and the AI credentials proxy still bill. A card is required, also for the 7-day trial.
Limits in those scenarios
Replicate does not fit the workday agent, the always-on agent or the 50-agent fleet as a work machine: it documents no persistent disk, and the fleet also needs Docker inside, which is not documented. Each Replicate prediction stops after 30 minutes unless support raises it, and Replicate publishes no disk or egress price. Paperspace fits all five scenarios and publishes no start limits; its short-job figures count run time only. On boat, the 1,000-job burst needs about 6,667 starts a day, above the 2,400 that the $2,000 plan includes, so extra starts are paid. Starts above 270 a minute fail with 429, and boat never queues a start.
| Scenario | Paperspace | Replicate | boat |
|---|---|---|---|
| One coding agent, workdays | None for this scenario |
| None for this scenario |
| One agent that never stops | None for this scenario |
| None for this scenario |
| 50 agents in parallel | None for this scenario |
| None for this scenario |
| Sandboxes for your users | None for this scenario |
| None for this scenario |
| 1,000 short jobs at once | None for this scenario |
|
|
Which one should you use?
Use Paperspace when
you want GPU or large CPU machines that a person sets up and uses over SSH, with a disk that stays, in the US, Canada or Amsterdam.
- A whole machine you log into. For a person who wants a lasting GPU box: every Paperspace machine is a full Linux VM with SSH, a kept disk and a public IPv4.
- Big CPU machines. For heavy single jobs: Paperspace CPU machines go up to 32 vCPU and 244 GB of memory.
- Machines in Amsterdam and Canada. For teams that need a region: Paperspace runs in New York, Amsterdam and Canada, while Replicate documents no region choice.
Use Replicate when
you want to call or serve an AI model on GPUs as an API, from its catalog or packaged with Cog, that scales with traffic and down to zero.
- Tens of thousands of models, one call away. For teams that want an image, video, speech or text model without hosting it: one API request runs it.
- Scale to zero, billed per second. For traffic that comes and goes: deployments scale down to zero and wake on the next prediction, billed per second.
- Newer GPUs, and Cog runs anywhere. For serving models: Replicate offers L40S, H100 and H200, and Cog images also run on your own servers.
Where boat fits
Paperspace and Replicate are GPU clouds, and boat has no GPUs, so model training and serving belong on them. For agents that need a machine of their own, boat gives each one a full VM that code can create, fork and drive. Your agent in boat can still call Replicate or use a Paperspace GPU when it needs one.
A machine code can drive
Paperspace's API cannot run commands inside a machine, so you build and keep an SSH layer. Replicate is reached through its prediction API, and each prediction stops at 30 minutes by default. boat runs commands, moves files and drives coding agents by API.
A disk saved every minute
Paperspace snapshots at most hourly, so a broken machine can lose up to an hour of work, and its disk bills while off. Replicate documents no persistent disk, so files that must last go to storage you set up. boat saves the disk every minute and forks running sandboxes; writes after the last good snapshot can be lost if a machine fails.
Docker, /dev/kvm and a desktop
Neither documents /dev/kvm, Replicate does not document Docker inside, and Paperspace streams a desktop only on Windows machines new users cannot create. Every boat sandbox has Docker, /dev/kvm and a 60 fps desktop with Chrome.
Use boat when you run agents that each need a machine of their own: a shell, root, Docker, /dev/kvm and a desktop, driven from code, with a disk saved every minute, within its per-plan start limits. Compare it one to one: boat vs Paperspace and boat vs Replicate.
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
Questions people ask
Is Paperspace cheaper than Replicate?
For CPU work in our five scenarios, yes. One agent that never stops is $63 against $237 a month, though Replicate does not fit that job as a work machine. 1,000 short jobs at once come to about $411 on Paperspace, counting run time only, against $3,580 on Replicate. For GPU eval jobs, Replicate comes to $1,338 against $1,454 a month.
Which is faster, Paperspace or Replicate?
Neither is in the public benchmarks we use, so we cannot say. boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB in StarSling's run of 3 October 2026, on a Ryzen 9 9950X host.
Can I run Docker on Paperspace or Replicate?
Docker runs inside a Paperspace machine, though nested virtualization is not documented. On Replicate, Cog builds your model into a Docker image, but running Docker inside it is not documented. A boat sandbox runs Docker, systemd and /dev/kvm like a normal Ubuntu server.
Which has better GPUs?
Paperspace rents Nvidia cards from the M4000 up to H100, and H100 needs a request. Replicate offers T4, L40S, A100 80GB, H100 and H200; H200 and most multi-GPU setups need a committed spend contract. boat has no GPU on any size.
Can I run an AI agent on Paperspace or Replicate?
On Paperspace you can run one on a VM you reach over SSH; you build your own command layer, copies come from templates of a stopped machine, and a Linux desktop is yours to install. Replicate runs models, not agents: it documents no shell or lasting disk, and each prediction stops at 30 minutes by default. boat is built for agents, and they can still call Replicate's API.
Where do Paperspace and Replicate run?
Paperspace runs machines in New York, Amsterdam and Canada. Replicate documents no region choice and lists its infrastructure subprocessors in the United States. boat runs on machines in Germany, Finland and France; you cannot pick a region, and snapshot copies can be stored outside the EU.
Try boat
Install the CLI, then run boat new.
7-day free trial