Nebius vs Baseten
Nebius is an AI cloud that rents full GPU and CPU virtual machines, Kubernetes and Slurm clusters, and you set up the rest yourself. Baseten is an inference cloud: you hand it a model or a Docker image and it runs replicas that scale with traffic, down to zero. Baseten launched Sandboxes for agents on November 5, 2026.
The companies and products
Nebius
Nebius is an AI cloud company, listed on Nasdaq and based in Amsterdam, that rents GPUs and the infrastructure around them.
You create a Compute VM from a preset (GPU or CPU only), add disks and an SSH key, and get a full Linux VM with root and an optional public IP. Around it Nebius sells Managed Kubernetes, Slurm clusters (Soperator), Serverless AI jobs and endpoints, object storage and networking. Serverless AI also runs Devlabs: managed coding environments from a container image (code-server, JupyterLab), with no agent or desktop built in. Nebius is built for teams that train, fine-tune and serve AI models, not as a sandbox for agents.
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. “42nd cheapest of 359” means 41 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 | Nebius: 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 on a Nebius VM once you install the agent and Docker yourself. Baseten Sandboxes take a custom image with the agent inside, but Docker inside is not documented. boat ships Docker and seven agent harnesses on every sandbox. | 214th of 365, with limits $2,272 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 | 252nd of 361, with limits $2,261 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 For short code bursts Nebius comes to $202 a month and Baseten's CPU instances to $1,856, because Baseten bills cold starts and, by default, 15 idle minutes per burst. Baseten Sandboxes are free until November 16, 2026, with no published price after that. | 115th of 336, with limits $202 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 300 user sandboxes are above both defaults: 2 VMs per region on Nebius and 10 Sandboxes per organization on Baseten, so either needs a quota raise first. boat runs 300 at once on the $100 plan with 2 seats. | 86th of 349, with limits $716 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 RL rollouts cost $3,049 a month on Nebius, $31,140 on Baseten CPU instances and $2,000 on boat in our model; none of the three is a full match. | 56th of 349, with limits $3,049 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 | 42nd cheapest of 359 $610 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 A Nebius CPU VM works as a CI runner once you install the tools, because you have root. Baseten documents no root or Docker in its model containers or its sandboxes. boat gives root and Docker on every sandbox, and a stopped sandbox has no compute charge between jobs. | 25th cheapest of 374 $297 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 A dev box needs a disk that stays. Nebius keeps disks on a stopped VM and bills them. Baseten Sandboxes keep files on standby but expire 7 days after creation by default, unless you change the policy or mount a volume. boat keeps a stopped disk for at least 14 days, and a saved snapshot keeps it longer. | 71st cheapest of 371 $91 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 Neither ships a desktop. You set it up: on Nebius, install a desktop, a VNC server and a browser on the VM; on Baseten, add a desktop and a VNC web client to a custom sandbox image and open it through a preview URL. boat streams a 60 fps desktop with Chrome from every sandbox. | 222nd of 367, with limits $610 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 | 271st of 356, with limits $305 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 | 294th of 343, with limits $302 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 | 174th of 327, with limits $6,049 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 | 217th of 351, with limits $156 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 rent GPUs for eval jobs, and boat has none. One H100 for this job costs $1,625 a month on Baseten serverless. On Nebius an on-demand H100 costs $1,127, but a fleet of 10 needs a quota raise first, because new accounts get 2 VMs per region. | 51st of 102, with limits $452 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 | Nebius | Baseten | boat |
|---|---|---|---|
| Kind of product | AI cloud: GPU and CPU VMs, Kubernetes, Slurm clusters, serverless jobs and Devlabs | Inference cloud for serving and training models; agent Sandboxes launched November 5, 2026, CLI labeled pre-release | Sandboxes for AI agents: VMs with agents, a desktop and hosting ready |
| Isolation | Every Compute VM is a full VM behind a hypervisor | 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, on a host dedicated to boat; vCPUs are shared |
| Root, Docker and nested virtualization | Root on your own VM, so you can install Docker; 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 L40S up to B300, on-demand or preemptible; GB200 and GB300 NVL72 through sales | Nvidia T4 up to B200, 1 to 8 per instance, multi-node over InfiniBand | None on any size |
| Biggest machine | CPU VMs up to 256 vCPU / 1,024 GB | Sandboxes up to 32 vCPU, 1 to 8 GiB per vCPU | 16 vCPU / 32 GB on the $100 plan and up |
| Files after a stop | Disks stay until you delete them and keep billing on their provisioned size ($0.071 per GiB-month) | Standby keeps files, memory and processes and bills only storage; sandboxes expire 7 days after creation by default; volumes keep files after deletion | Disk saved every minute and on stop; no compute charge while stopped; disk kept at least 14 days (save a snapshot to keep it longer) |
| Snapshots and fork | Disk snapshots of a running VM; you set up any schedule yourself; to fork, make a disk from the snapshot, then a VM from the disk | No fork documented; a new sandbox starts from the image, and a volume mounts on one sandbox at a time | Disk saved every minute and on stop; fork a running sandbox from its last 1-minute snapshot in one call |
| Scale to zero and wake on request | No wake on request; a VM runs until you stop it | Deployments scale to zero and start on the next request; standby sandboxes resume with memory on the next request | No wake on request; you resume a stopped sandbox through the API, and a resume is like a reboot |
| Billing | Per second for vCPU and RAM, no plan fee; disks billed separately, also while stopped | CPU instances per minute, rounded up, including image builds and cold starts, no plan fee on Basic; Sandboxes free until November 16, 2026, later price not published | Plans from $20 a month; the plan price comes back as VM time for that month, spent per second; extra starts $0.03 to $0.07 |
| Desktop and computer use | No desktop product; you set it up: install a desktop, a VNC server and a browser with root | None documented; you set it up: add a desktop and a VNC web client to a custom sandbox image and open it through a preview URL | Built-in Linux desktop and Chrome at 1920x1080, 60 fps, with computer tools for agents |
| Coding agents | None built in; you install agents over SSH; Devlabs give browser IDEs with no agent | You install the agent in a custom sandbox image or drive the sandbox through the CLI or its MCP endpoint; Baseten Switch routes Claude Code or Codex to its Model APIs | Seven harnesses installed, including Claude Code, Codex and OpenCode, with an API to prompt, steer and stream events |
| Reaching the machine | Optional public IPv4 per VM, SSH and Nebius Tunnels; you set up TLS yourself | HTTPS API endpoint per deployment; sandbox previews give public or private HTTPS URLs per port | SSH over IPv4 through a shared host and port; up to 50 HTTPS URLs, token-gated by default; public IPv6; no dedicated IPv4 |
| Outbound network rules | VPC security groups allow or deny egress by IP range, port and protocol | Deployments: TCP 443 and 80 by default; IP and domain allowlists once enabled for your organization; not documented for sandboxes | Open internet on every plan; no provider-side allowlist |
| Keeping keys out of the machine | SecretStash stores secrets that a VM fetches with its service account, so the secret reaches the VM | Secrets are passed into the model as arguments or files under /secrets | Optional AI credentials proxy ($1 a day it is on, plus $1 per 1,000 requests) for model providers and HTTPS APIs you register |
Every feature Battleships tracks (43)
| Feature | Nebius | 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 | Yes | 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 | Yes | No documented way | Yes |
| Live fork (no pause) | Yes | 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 | No documented way | Yes | No documented way |
| Wake on request | No documented way | Yes | No documented way |
| HTTPS preview URLs | Yes | Yes | Yes |
| Raw TCP inbound | Not published | Not published | Yes |
| Public IPv4 | Yes | No documented way | No documented way |
| Egress allowlist | Yes | 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 | Not published | No documented way | Yes |
| Usage per sandbox (API) | Yes | 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
Nebius and Baseten are not in the public benchmarks we use, so we cannot say which 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. The boat fleet also has Ryzen 7 9700X hosts, and boat does not promise a CPU model. Nebius CPU VMs run on AMD EPYC Genoa or Intel Ice Lake, and Baseten publishes no CPU benchmark.
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. Nebius 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 | Nebius, 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 | $25 · Pay as you go | $34 · Basic · with limits | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $91 · Pay as you go | $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 | $1,127 · Pay as you go | $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 | $305 · Pay as you go | $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 | $610 · Pay as you go | $6,228 · Basic · with limits | $8,800 · $2000/mo · with limits |
Nebius: Nebius bills running VMs per second for vCPU and RAM with no plan fee, and bills disks at $0.071 per GiB-month on their provisioned size, also while the VM is stopped. The scenarios use its lowest CPU rate from 1 October 2026: Intel Ice Lake in Finland, $0.012 per vCPU-hour plus $0.0045 per GiB-hour. Its smallest 4 vCPU VM has 16 GiB, twice boat's 8 GB.
Baseten: Baseten Sandboxes are free until November 16, 2026, and their later price is not published, so the scenarios use CPU instances. These bill per minute, rounded up, from image build and cold start until the replica stops; 4 vCPU / 16 GiB is $0.00346 a minute. Basic has no plan fee.
Baseten: Plans cost $20, $100, $500 or $2,000 a month, and a card is required, also for the 7-day trial. The whole plan price comes back as VM time for that month and does not carry over. A 4 vCPU / 8 GB sandbox costs $0.036 an hour, billed per second while it runs. A stopped sandbox has no compute charge and keeps its disk for at least 14 days; extra saved snapshots and the AI credentials proxy still charge. Above the included starts (200 a day on $20), each extra start costs $0.03 to $0.07.
Limits in those scenarios
Nebius starts each region at 2 regular VMs and, in most regions, 200 vCPU across CPU-only VMs (0 in some regions), so 50 agents, 300 user sandboxes and 1,000 short jobs at once all need a quota raise first. Baseten's CPU instances keep no files after a stop, so the table marks them as not fitting. Its Sandboxes keep files but run 10 at once per organization by default and have no published price after November 16, 2026. A Baseten monthly budget never stops dedicated deployments. boat runs 100 to 2,000 sandboxes at once by plan, but the 1,000-job burst needs more starts than the $2,000 plan includes, and starts above 270 a minute fail with 429. boat keeps a stopped disk for at least 14 days unless you save a snapshot.
| Scenario | Nebius | 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 Nebius when
you want full GPU or CPU VMs with root, very large machines, Slurm or Kubernetes clusters, IP-level egress rules, a SOC 2 Type II report, or a region in the UK, Israel or the US.
- Full VMs with root and a disk that stays. Useful when you need a full Linux machine: Nebius gives root on a VM whose disks stay after a stop until you delete them.
- Very large machines and GPU clusters. Useful for big training or batch jobs: CPU VMs reach 256 vCPU, and Nebius runs Slurm and Kubernetes clusters on its GPUs.
- Network rules, more regions and audit reports. Useful if you need egress rules by IP range, a region in Spain, the UK, Israel or the US, or a compliance report.
Use Baseten when
you serve or train models and want deployments that scale to zero and wake on request, sandboxes that keep memory on standby, hosted open models per token, or your own cloud account.
- Scale to zero, wake on request. Useful for spiky traffic: deployments scale to zero when idle, and standby sandboxes resume with memory on the next request.
- Model serving done for you. Useful if you ship a model, not a server: package it with Truss or a Docker image, or call hosted open models per token.
- Your own cloud account. Useful on Enterprise when the workload must run in your own AWS, GCP or OCI account.
Where boat fits
Nebius and Baseten are GPU clouds first. Baseten's Sandboxes are new, and on Nebius you build the agent setup yourself. If you run agents that write code, use a browser or run for days, boat gives each one a full VM with the agents, a desktop and a disk saved every minute. Keep GPU training and model serving on Nebius or Baseten; boat has no GPU, runs only in Europe and limits starts per plan.
Coding agents installed, driven from outside
On Nebius you install every agent over SSH. On Baseten you build the agent into a custom sandbox image or drive the sandbox from outside. boat installs seven harnesses, including Claude Code, Codex and OpenCode, with a prompt, steer and events API.
Disk saved every minute, fork in one call
Nebius snapshots run when you call them, and stopped VMs still pay for disks. Baseten Sandboxes keep memory on standby but document no fork. boat saves the disk every minute, forks a running sandbox from its last snapshot, and a stopped sandbox has no compute charge.
A desktop and keys kept off the machine
On both, a computer-use agent means you install a desktop, VNC and a browser. Both also put the real secret on the machine. boat streams a 60 fps desktop, and its optional paid credentials proxy keeps AI keys outside.
Use boat when you run agents: full VMs with root and Docker, seven coding agents and a 60 fps desktop installed, the disk saved every minute, forks of a running sandbox, and no compute charge while stopped. Compare it one to one: boat vs Nebius 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 Nebius cheaper than Baseten?
Against Baseten's CPU instances, yes, in every scenario we priced. One agent running all month is $91 on Nebius and $136 on Baseten, and 50 agents in parallel are $1,127 against $1,701. Baseten's CPU instances keep no files after a stop. Baseten Sandboxes are free until November 16, 2026, so until then they cost less than both. boat costs $26 for the agent all month and $317 for 50 agents.
Which is faster, Nebius or Baseten?
Neither is in the public benchmarks we use, so we cannot say. In StarSling's run of 3 October 2026, boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, on a Ryzen 9 9950X host.
Can Nebius or Baseten run Docker?
On Nebius you get root on a full VM, so you can install Docker. Baseten runs your Docker image as a model server and builds sandbox images from a Dockerfile, but it does not document Docker inside its containers or sandboxes. Every boat sandbox has root and Docker on its own kernel.
Does Baseten keep files like a Nebius VM?
Baseten Sandboxes keep files, memory and processes on standby, and volumes keep files after deletion. A sandbox expires 7 days after creation by default, which you can change. Baseten model replicas start fresh from the image. Nebius disks stay until you delete them and keep billing while the VM is stopped. A stopped boat sandbox has no compute charge and keeps its disk for at least 14 days.
Does boat have GPUs like Nebius and Baseten?
No. boat runs Linux VMs on AMD Ryzen hosts and has no GPU on any size. Train or serve models on Nebius or Baseten and let your agents work on boat.
Is boat SOC 2 compliant?
SOC 2 is in progress, and boat has no SOC 2 report today. Nebius has a SOC 2 Type II report and ISO/IEC 27001; Baseten says it is SOC 2 Type II certified. Live boat status is at status.ascii.dev and the data retention policy at docs.boat.dev/data-retention.
Try boat
Install the CLI, then run boat new.
7-day free trial