Paperspace vs Baseten
Paperspace, part of DigitalOcean, rents whole Linux VMs, with or without Nvidia GPUs, by the hour, and you log into them over SSH. Baseten runs your models as containers on its GPUs that scale with traffic, billed per minute. Baseten launched agent Sandboxes on November 5, 2026.
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
Baseten
Baseten is an inference cloud: it runs AI models on GPUs for production apps, sells hosted open models and training, and now runs agent sandboxes.
You package a model with Truss, its open-source format, or bring a Docker image, and Baseten runs it as replicas on Kubernetes that scale up with traffic and down to zero. It also sells Model APIs (open models priced per token) and training jobs on the same GPUs. Sandboxes, isolated environments where an agent runs commands and works with files, launched on November 5, 2026. They are free until November 16, 2026, and later pricing is not published. This page compares boat with both the CPU instances and the Sandboxes.
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 | Baseten: price rank | boat: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A coding agent fits better on a Paperspace VM than in a Baseten model container, but you set up each machine yourself over SSH. boat ranks 8th in our default preset with agents and Docker ready. | 164th of 365, with limits $1,169 a month | 234th of 365, with limits $2,619 a month | 5th cheapest of 365 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB | 229th of 361, with limits $1,560 a month | 278th of 361, with limits $3,258 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 are a compromise on all three: Paperspace bills by the hour and does not say whether a part hour rounds up, Baseten CPU instances bill cold starts and, by default, 15 idle minutes, and boat limits starts per plan. | 71st of 336, with limits $61 a month | 313th of 336, with limits $1,856 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's API cannot run code inside a machine, so you build that layer over SSH. Baseten Sandboxes run 10 at once per organization by default, and support raises the limit. boat ranks 2nd in our default preset. | 63rd of 349, with limits $428 a month | 273rd of 349, with limits $1,868 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 | 315th of 349, with limits $31,140 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 In our default agent evals preset, boat ranks 8th and Paperspace 13th. Baseten's CPU instances are a compromise because they keep no disk; its Sandboxes are not priced. | 22nd cheapest of 359 $331 a month | 297th of 359, with limits $1,609 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 In our default CI preset, boat ranks 2nd and Paperspace 7th; a Paperspace VM runs Docker inside. Baseten documents neither root nor Docker for its model containers or its sandboxes. | 12th cheapest of 374 $205 a month | 323rd of 374, with limits $1,119 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, since its VM keeps a disk. Baseten's CPU instances lose their files when a replica stops, and its Sandboxes have no published price after November 16, 2026. boat ranks 13th in our default preset at $26 a month, Paperspace 35th at $63. | 50th cheapest of 371 $63 a month | 186th of 371, with limits $130 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 that new users cannot create; on Linux you install one yourself. On Baseten you add a desktop and a VNC web client to a custom sandbox image. boat streams a 60 fps desktop with Chrome and computer-use tools. | 171st of 367, with limits $298 a month | 314th of 367, with limits $1,799 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 | 331st of 356, with limits $1,194 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 | 275th of 343, with limits $183 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 | 128th of 327, with limits $3,654 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 | 275th of 351, with limits $381 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 Paperspace ranks 20th and Baseten 21st on H100. boat has no GPU on any size. | 23rd cheapest of 102 $1,454 a month | 25th cheapest of 102 $1,625 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 | Baseten | boat |
|---|---|---|---|
| Kind of product | GPU and CPU VM cloud for machine learning; DigitalOcean points new users to GPU Droplets | Inference cloud for serving and training models; agent Sandboxes launched November 5, 2026, CLI labeled pre-release | Sandbox cloud for agents: VMs you create, fork and drive from code |
| Isolation | Every machine is a full VM with its own kernel | Sandboxes: a microVM per sandbox with its own kernel, per launch coverage. Model deployments: containers | Every sandbox is a full VM with its own kernel |
| Root, 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 | Model deployments run as the non-root app user; the sandbox docs do not document root, Docker or /dev/kvm | Docker, root, systemd and /dev/kvm on every sandbox |
| GPUs | Nvidia M4000 up to H100, up to 8 per machine; H100 needs a request | Nvidia T4 up to B200, 1 to 8 per instance, multi-node over InfiniBand | No GPU on any size |
| Files after a stop | Disk kept while the machine exists, billed while off; scheduled snapshots at most hourly, up to 10 kept | Model replicas start fresh from the image. Sandboxes on standby keep files, memory and processes and bill only storage; they expire 7 days after creation by default (configurable); volumes keep files after deletion | 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 made from a machine that is turned off | No fork documented; you start a new sandbox from the image, and a volume mounts on one sandbox at a time | Fork a running sandbox in one call, from its last 1-minute snapshot |
| Working inside from code | The API and CLI manage machines but cannot connect to them; you run commands over SSH from your own code | Deploy a model and call its endpoint; Sandboxes run commands and file operations through the CLI or an MCP endpoint | API, SDK and CLI to run commands, move files and fork |
| Scale to zero and wake on request | No wake on request; machines can auto-shut down after a set time | Deployments scale to zero and start on the next request; standby sandboxes resume with memory and processes | No wake on request; auto-stop runs on a timer (default 1 hour), and you resume a stopped sandbox through the API |
| Coding agents | No agent templates; ML templates such as ML-in-a-Box. You install agents yourself | You install the agent in a custom sandbox image, or drive the sandbox from outside; Baseten Switch routes Claude Code or Codex to Model APIs | Seven coding-agent harnesses installed (Claude Code, Codex, OpenCode and others), with a prompt, steer and events API |
| Desktop and computer use | Built-in streaming only on Windows machines, closed to new users; on Linux you install a desktop and VNC yourself | None documented; you add a desktop and a VNC web client to a custom sandbox image and open it through a preview URL | Built-in 1920x1080, 60 fps desktop and Chrome stream with computer-use tools; embeddable stream |
| Reaching the machine | A public IPv4 on every Linux machine, any port, with a static option; no managed HTTPS URLs | HTTPS API endpoint per deployment; sandbox previews give public or private HTTPS URLs per port | Up to 50 HTTPS URLs per sandbox, token-gated by default; public IPv6; the IP changes at each resume |
| Keeping keys out of the machine | No credentials proxy; Gradient secrets are injected as environment variables | Secrets are passed into the model as arguments or files under /secrets | 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 | US and EU for GPU deployments; self-hosted or hybrid in your own cloud account on Enterprise | Germany, Finland and France; no US or Asia region, and you cannot pick one. Snapshot copies can be stored outside the EU |
| Billing step | Hourly rates while on; a finer billing step is not published. Disk and public IP bill while off | Per minute for CPU and GPU instances, including image builds and cold starts; Sandboxes free until November 16, 2026, later price not published | Per second, from a monthly plan ($20 minimum) returned as VM time for that month |
Every feature Battleships tracks (43)
| Feature | Paperspace | Baseten | 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 | Yes | 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 | Yes | 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 | Yes | No documented way |
| Open source | No documented way | No documented way | No documented way |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | Yes | No documented way | Yes |
| Usage per sandbox (API) | No documented way | Yes | Yes |
| Audit logs | Not published | Not published | No documented way |
| SSO | Yes | Yes | No documented way |
| SOC 2 | Yes | Yes | No documented way |
| HIPAA | Yes | Yes | No documented way |
| EU data residency | Yes | Yes | No documented way |
| Zero data retention | Not published | Not published | Yes |
Performance
Paperspace and Baseten are not in the public benchmarks we use, so we cannot say which of the three is faster. 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.
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 Baseten 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 | Baseten, 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 | $34 · Basic · 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 | $136 · Basic · 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 | $1,701 · Basic · 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 | $1,246 · Basic · with limits | $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 | $6,228 · Basic · with limits | $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. The boot disk (from $5 a month for 50 GB) and the public IP (up to $3 a month) bill until you delete the machine. There is no plan fee.
Baseten: Baseten CPU instances bill per minute, rounding partial minutes up, from image build and cold start until the replica stops. There is no 4 vCPU / 8 GiB CPU instance, so the scenarios use 4 vCPU / 16 GiB at $0.00346 a minute. There is no plan fee on Basic. Sandboxes are free until November 16, 2026, and their later price is not published.
Baseten: 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
Baseten's CPU instances are model containers with no persistent disk, so files are lost when a replica stops, and the table marks them as not fitting. Baseten Sandboxes keep files on standby but have no published price after November 16, 2026, so the table does not price them; they run 10 at once per organization by default. Paperspace fits every scenario and publishes no start or machine count limits. 1,000 short jobs at once come to about $411 a month there, counting run time only, against $6,228 on Baseten CPU instances. On boat that 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 | Baseten | 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 a whole GPU or CPU VM with a kept disk and a public IP that a person sets up and uses over SSH, in New York, Canada or Amsterdam, with no published start limits.
- A whole VM you log into. Useful if a person works on the machine: Paperspace gives a full Linux VM with a kept disk and its own public IPv4.
- Big CPU machines. Useful for heavy single jobs: Paperspace CPU machines go up to 32 vCPU and 244 GB of memory.
- Machines in New York and Canada. Useful if your users or data are in North America: Paperspace runs in New York and Canada as well as Amsterdam.
Use Baseten when
you serve or train models in production and want deployments that scale to zero, wake on request, and can run in your own cloud account, or sandboxes that keep memory on standby.
- Models served for you. Useful if your product serves a model: Baseten runs it as replicas that scale with traffic, on GPUs up to B200.
- Scale to zero, wake on request, memory kept on standby. Useful for spiky traffic and idle agents: deployments scale to zero, and Sandboxes resume with memory and processes on the next request.
- Your own cloud and egress allowlists. Useful for strict networks: on Enterprise Baseten can run in your own cloud account, with per-deployment egress allowlists.
Where boat fits
Paperspace and Baseten are GPU clouds, and boat has no GPUs, so keep model training and serving on one of them. For agents that write code, test it and use a browser, boat gives each one a full VM with root, Docker, a desktop and coding agents installed. Baseten's new Sandboxes also target agents, with memory kept on standby and no published price after November 16, 2026.
Driven from code, with agents inside
Paperspace's API cannot run commands inside a machine, so you build and keep your own SSH layer. Baseten Sandboxes run commands from outside, and you add agents to a custom image. boat has Claude Code, Codex and OpenCode installed, driven from outside by API.
Root and Docker in a VM that keeps its disk
Baseten model containers run as a non-root user and lose their files when they stop, and its sandbox docs do not cover root or Docker. Paperspace keeps the disk but snapshots at most hourly and bills it while off. boat saves the disk every minute and has no compute charge while stopped.
Fork and a desktop for computer use
Copying a Paperspace machine means turning it off and making a template; Baseten documents no fork. On both, a Linux desktop is yours to install. boat forks a running sandbox in one call and streams a 60 fps desktop with Chrome.
Use boat when the work is agents rather than models: full VMs with root and Docker, coding agents and a desktop inside, snapshots every minute and forks of a running sandbox, within its per-plan start limits. Compare it one to one: boat vs Paperspace and boat vs Baseten.
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 Baseten?
For CPU work in the scenarios we priced, yes. One agent that never stops is $63 a month on Paperspace and $136 on Baseten CPU instances, and 50 agents in parallel are $734 versus $1,701. For 1,000 short jobs at once, Paperspace comes to about $411 against $6,228. On boat the first two cost $26 and $317. Baseten Sandboxes are free until November 16, 2026, so until then they cost less than all three.
Which is better for GPU work?
Both rent Nvidia GPUs, up to 8 per machine. Baseten has newer cards such as B200 and scales model servers with traffic, while Paperspace gives you a whole VM with a GPU that you log into. In our default GPU eval jobs preset, Paperspace ranks 20th and Baseten 21st. boat has no GPU on any size.
Which is faster?
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 Paperspace or Baseten run Docker?
A Paperspace machine is a full VM, and Docker runs inside it. Baseten can run your own Docker image as a model server, but its model containers run as a non-root user, and its docs do not cover Docker inside a container or a sandbox. Every boat sandbox has root, Docker and /dev/kvm.
Does either have sandboxes for AI agents?
Baseten launched Sandboxes on November 5, 2026. They run 10 at once per organization by default, keep memory and files on standby, and expire 7 days after creation by default, which you can change. They are free until November 16, 2026; later pricing is not published. Paperspace rents VMs that you reach over SSH. boat is a sandbox cloud for agents on every plan.
Where does my data run?
Paperspace runs in New York, Amsterdam and Canada. Baseten runs GPU deployments in the US or EU, and on Enterprise in your own cloud account. 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