Hyperstack vs Replicate

Hyperstack rents you whole GPU and CPU virtual machines with root and SSH, billed by the minute. Replicate runs AI models as containers behind a prediction API, on hardware that scales up with requests and down to zero.

The companies and products

Hyperstack

Hyperstack is the GPU cloud of NexGen Cloud, a UK company founded in 2020: you rent GPU and CPU virtual machines, Kubernetes clusters and storage by the minute.

A Hyperstack VM is a full virtual machine with root and SSH, booted from an Ubuntu image (with or without Docker) and sized from a fixed list of flavors, from one RTX A4000 to 8-GPU B300 nodes, plus CPU-only flavors from 4 to 32 vCPUs. It also runs managed Kubernetes, volumes, object storage and an AI Studio inference API. It is built for teams that train, fine-tune and serve models, not as a sandbox API for agents, so this page compares the parts that overlap: VMs as machines for agent work.

Company: NexGen Cloud, founded in 2020, head office in London
Kind of product: GPU cloud: GPU and CPU VMs, Kubernetes clusters, volumes, object storage, AI Studio inference
Isolation: Full virtual machines; all flavors map 1:1 to physical resources with no oversubscription
Regions: NORWAY-1 (Vestland, Norway), CANADA-1 (Quebec), US-1 (Texas)
Billing: Prepaid balance billed per minute; stopped (SHUTOFF) VMs bill at the full rate, hibernated VMs bill only disk and a kept IP
Compliance: NexGen Cloud says it has achieved SOC 2 Type 2; data centers are Tier 3 certified

Replicate

Replicate runs AI models with an API: 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, 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.

Company: Founded 2019, Y Combinator Winter 2020, San Francisco; joined Cloudflare in November 2025 and continues as its own brand with the same API
Product: Model inference API: public models, fine-tunes, and custom Cog models run as autoscaling deployments
Isolation: Each model runs as a Docker container built by Cog; the docs do not name the isolation layer
Hardware: CPU (1 or 4 vCPU), T4, L40S, A100 80GB, H100 and H200, up to 8 GPUs per instance; most multi-GPU and all H200 need a committed spend contract
Run time: Each prediction times out after 30 minutes unless support raises it
Hosting: Infrastructure subprocessors (AWS, CoreWeave, GCP) listed in the United States; no region choice documented
Open source: Cog is open source and its images run anywhere Docker runs; the hosted platform is not

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. “122nd cheapest of 377” means 121 of the other 376 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	Hyperstack: price rank	Replicate: price rank	bochka: price rank
Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A software factory needs root, Docker and a lasting disk: Hyperstack has them but bills a stopped VM in full, and Replicate keeps no disk at all.	343rd of 367, with limits $11,458 a month	268th of 367, with limits $4,340 a month	3rd cheapest of 367 $634 a month
Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB	337th of 366, with limits $12,775 a month	335th of 366, with limits $11,300 a month	5th cheapest of 366 $657 a month
Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short code runs suit Replicate's model, where a request starts work on an instance that bills per second and scales back to zero.	172nd of 345, with limits $1,167 a month	258th of 345, with limits $295 a month	186th of 345, 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 Replicate fits 300 user sessions at once; Hyperstack does not publish how many VMs can run at once, so that is not confirmed there.	173rd of 361, with limits $4,207 a month	317th of 361, with limits $3,705 a month	Cheapest of 361 $209 a month
RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month	173rd of 360, with limits $17,500 a month	174th of 360, with limits $17,900 a month	34th of 360, with limits $2,000 a month
Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month Replicate fits agent evals as short, separate runs that bill per second, while Hyperstack is a compromise here; bochka fits with a full VM per eval.	204th of 364, with limits $3,250 a month	67th cheapest of 364 $1,440 a month	10th cheapest of 364 $200 a month
CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI fits Hyperstack, where each job gets a full VM with Docker, but not Replicate, which has no shell and no documented Docker inside.	122nd cheapest of 377 $1,560 a month	311th of 377, with limits $864 a month	4th cheapest of 377 $86 a month
24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box needs SSH and a disk that keeps your installs: Hyperstack has both, Replicate has neither, and bochka adds free stops.	115th cheapest of 368 $475 a month	306th of 368, with limits $226 a month	18th cheapest of 368 $26 a month
Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Neither gives an agent a screen: Hyperstack offers SSH and text console logs, Replicate has no desktop, and bochka streams a 60fps desktop.	332nd of 369, with limits $3,250 a month	263rd of 369, with limits $1,296 a month	6th cheapest of 369 $200 a month
Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month	335th of 367, with limits $1,750 a month	322nd of 367, with limits $1,296 a month	7th cheapest of 367 $200 a month
Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month	327th of 346, with limits $1,750 a month	272nd of 346, with limits $439 a month	6th cheapest of 346 $90 a month
Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month	310th of 338, with limits $35,000 a month	221st of 338, with limits $8,775 a month	71st of 338, with limits $1,800 a month
Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month	309th of 360, with limits $876 a month	306th of 360, with limits $855 a month	2nd cheapest of 360 $200 a month
GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month Both rent Nvidia GPUs and fit this job; Hyperstack ranks 2nd and Replicate ranks 22nd of 98, and bochka has no GPUs.	2nd cheapest of 98 $625 a month	22nd cheapest of 98 $1,338 a month	Can't run it
macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month	Can't run it	Can't run it	Can't run it

Key features

Feature	Hyperstack	Replicate	bochka
What you get	A full VM with root and SSH, from a fixed list of GPU and CPU flavors	A model packaged as a Cog container and called through a prediction API	A full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts
GPUs	Nvidia RTX A4000 to H100, H200, B200 and B300, up to 8-GPU nodes, on demand, spot or reserved	T4, L40S, A100 80GB, H100 and H200, up to 8 per instance; most multi-GPU and all H200 need a committed spend contract	None, CPU only
Ready-made models	An AI Studio inference API next to the VMs	Thousands of public and official models for images, video, speech and text, plus fine-tuning	None; you bring the model API you call
Shell access	Root and SSH	No SSH; you call the model through the HTTP API or client libraries	SSH, root and systemd, like a normal Ubuntu server
Docker	Ubuntu 22.04 and 24.04 images with or without Docker	The model is the container; Docker inside is not documented	Docker, Docker Compose and /dev/kvm on every machine
Files after a stop	The whole disk stays; a stopped VM bills at the full rate, a hibernated one bills only disk and a kept IP	No persistent disk; instances start fresh, and API inputs and outputs are deleted after an hour by default	The whole disk is saved every minute and on stop, and a stopped sandbox costs nothing
Snapshots and fork	Snapshots on request with the VM shut off, up to 100 GB; no fork	No snapshots or fork; you add replicas of the same image	Automatic snapshots every minute, and fork of a running sandbox
Scale to zero	No; VMs never stop on their own, you hibernate them by hand	Yes; deployments scale to zero and start again on the next prediction	No; a stopped sandbox is started by API
Run time	No maximum runtime documented	Each prediction times out after 30 minutes unless support raises it	No session cap on paid plans; you set the time to live
Reaching the machine	A public IPv4 you can keep static; you open ports and run your own TLS	One prediction API per model behind API token auth; no own IP or ports	HTTPS URLs for up to 50 ports, token-protected; raw TCP; dedicated IPv4 or IPv6
Network rules	Per-VM firewall: inbound closed by default, outbound rules by IP, protocol and port	No firewall or outbound rules offered	Open internet out; inbound through HTTPS URLs and TCP ports
Desktop and browser	SSH and text console logs; no graphical desktop	No desktop or browser product	60fps Linux desktop with Chromium, computer use and browser control
Keeping keys out of the machine	No credentials proxy; keys go in with cloud-init, files or environment variables	Secrets are sent to the model as inputs and redacted from logs afterwards	Credentials proxy for model providers and any HTTPS API you register
Regions	Norway, Canada and the US	Infrastructure subprocessors listed in the US; no region choice documented	EU only: Germany, Finland and France

Every feature Battleships tracks (43)

Feature	Hyperstack	Replicate	bochka
Full VM (own kernel)	Yes	No	Yes
Docker inside	Yes	No	Yes
Nested virtualization	Not published	No	Yes
Root	Yes	Not published	Yes
Systemd	Not published	Not published	Yes
Desktop GUI	No	No	Yes
Browser	No	No	Yes
Computer-use API	No	No	Yes
SSH	Yes	No	Yes
Persistent disk	Yes	No	Yes
Pause/resume	Yes	No	Yes
Automatic snapshots	No	No	Yes
Fork/clone	No	No	Yes
Live fork (no pause)	No	No	Yes
Memory fork	No	No	No
Custom image (Docker or snapshot)	Yes	Yes	Yes
≥24 h sessions	Yes	Yes	Yes
Idle auto-stop	No	Yes	No
Wake on request	No	Yes	No
HTTPS preview URLs	No	Yes	Yes
Raw TCP inbound	No	Not published	Yes
Public IPv4	Yes	No	No
Egress allowlist	Yes	No	No
Secret proxy	No	No	Yes
Secret proxy for any API	No	No	Yes
Preinstalled agents	No	No	Yes
Agent harness API	No	No	Yes
MCP server	Yes	Yes	No
GPU	Yes	Yes	No
Windows	No	No	No
Arm64	No	No	No
Self-hosting	No	Yes	No
BYOC	No	No	No
Open source	No	Yes	No
Scoped API keys	Not published	Not published	Yes
Spend limits	Not published	Yes	Yes
Usage per sandbox (API)	No	No	Yes
Audit logs	Not published	Yes	No
SSO	No	No	No
SOC 2	Yes	Not published	No
HIPAA	No	Not published	No
EU data residency	Yes	No	Yes
Zero data retention	Not published	Not published	Yes

Performance

HyperstackNot measured

ReplicateNot measured

bochka26.69 runs/s

There is no comparable CPU measurement for Hyperstack or Replicate: neither is in the StarSling hpc-sandbox-benchmarks runs we use, so we cannot say which is faster. bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB in the run of 2026-09-29, on dedicated AMD Ryzen 9 9950X hosts. Both Hyperstack and Replicate pick their hardware for GPU work, which this test does not cover.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Hyperstack and Replicate have no published run in this suite yet, so there is no like-for-like number to compare.

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.

Scenario	Hyperstack, a month	Replicate, a month	bochka, 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	$114 · Prepaid on-demand	$57 · Usage · with limits	$20 · $20/mo
One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month	$475 · Prepaid on-demand	$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	$5,720 · Prepaid on-demand	$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	$1,750 · Prepaid on-demand · with limits	$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	$3,500 · Prepaid on-demand · with limits	$3,580 · Usage	$2,000 · $2000/mo · with limits

Hyperstack: Hyperstack bills a running VM per minute from a prepaid balance, at a fixed hourly price per flavor, with no charge for bandwidth. CPU flavors start at 4 vCPU / 4 GB, and a stopped VM keeps billing in full, so the totals assume you hibernate it between runs.

Replicate: Replicate bills per second by hardware tier, from prepaid credit that expires after a year. Public models bill only while they run; private models and deployments also bill setup and idle time.

Replicate: Plans start at $20 a month, and the whole fee becomes usage credit spent per second on running sandboxes. A stopped sandbox costs nothing and keeps its whole disk.

Limits in those scenarios

Replicate does not fit the workday agent, the always-on agent or the 50-agent fleet, because it has no persistent disk, and the fleet also needs Docker inside, which is not documented. Hyperstack fits those three but does not publish how many VMs can run at once, so 300 user sandboxes and 1,000 short jobs at once are not confirmed there, while Replicate fits both. Each Replicate prediction stops after 30 minutes unless support raises it. bochka's $2,000 plan queues the 1,000-job burst because it needs more starts per day than the plan allows.

Scenario	Hyperstack	Replicate	bochka
One coding agent, workdays	None that bite	missing persistent disk no persistent disk: files lost when it stops or sleeps disk beyond 0 GiB: price unknown egress price not published	None that bite
One agent that never stops	None that bite	missing persistent disk no persistent disk: files lost when it stops or sleeps disk beyond 0 GiB: price unknown egress price not published	None that bite
50 agents in parallel	how many machines can run at once: limit not published, probably fine at 50 (the typical published limit is 50)	missing persistent disk, Docker inside no persistent disk: files lost when it stops or sleeps disk beyond 0 GiB: price unknown egress price not published	None that bite
Sandboxes for your users	not confirmed: how many machines can run at once (you need 300; the typical published limit is 50)	disk beyond 0 GiB: price unknown egress price not published	None that bite
1,000 short jobs at once	not confirmed: how many machines can run at once (you need 1,000; the typical published limit is 50)	disk beyond 0 GiB: price unknown egress price not published	max 2,400 starts/day (needs ~6,667) max 600 starts/hour (needs ~1,000) bursts above 90 starts/min are queued (a burst of 333 takes ~4 min to start)

Which one should you use?

Use Hyperstack when

you want whole GPU or CPU machines with root, SSH and a lasting disk, firewall rules, or regions in Canada and the US.

Whole GPU machines you control. Useful if you want root and SSH on a GPU box: Hyperstack rents full VMs from one RTX A4000 to 8-GPU B300 nodes.
A disk that stays between runs. Useful if your work builds up files: a Hyperstack VM keeps its whole disk, and hibernation keeps it while compute billing stops.
Firewall rules and North American regions. Useful if you must limit outbound traffic by IP and port, or keep data in Canada or the US.

Use Replicate when

you want to call or serve an AI model as an API, from its catalog or packaged with Cog, that scales with traffic and down to zero.

Thousands of models, one API call away. Useful if you want to call an image, video, speech or language model without hosting it yourself.
Scale to zero and wake on request. Useful for models that wait for traffic: deployments scale down to zero and start again on the next prediction.
Cog images run anywhere. Useful if you want an exit: Cog, the open-source packaging tool, builds standard Docker images you can run on your own servers.

Where bochka fits

Hyperstack and Replicate are GPU clouds for training and serving models, and bochka has no GPUs, so model work belongs on them. If what you are really running is agents that write and run code for hours, look at bochka: each agent gets a full VM with Docker, a disk saved every minute, a desktop and coding agents built in. Your agents in bochka can still call a model on Replicate or a GPU VM on Hyperstack.

A disk saved every minute, free stops and fork

On Hyperstack a stopped VM bills at the full rate unless you hibernate it, and snapshots need the VM shut off. On Replicate an agent's files are gone when the instance scales down, so every run sets itself up again.

Coding agents driven from outside, keys kept out

On both you install and run the agent yourself, and keys sit inside as files, variables or model inputs where code the agent runs can read them. bochka preinstalls Claude Code, Codex and more, and keeps real keys in a credentials proxy.

A desktop, Docker and HTTPS URLs on every machine

Hyperstack gives SSH and a public IP, so a desktop, VNC server and certificates are yours to set up on every image. Replicate has no desktop, no ports and no documented Docker inside, and stops each prediction at 30 minutes.

Use bochka when your agents need a machine of their own for hours: a shell, root, Docker, a browser and a desktop, on a full VM whose disk is saved every minute and costs nothing while stopped. Compare it one to one: bochka vs Hyperstack and bochka vs Replicate.

Questions people ask

Is Hyperstack cheaper than Replicate?

For GPU eval jobs, yes: about $625 a month on Hyperstack against $1,338 on Replicate. For CPU agent work Replicate's totals are lower, such as $237 against $475 a month for one agent that never stops, but Replicate does not fit that case because it keeps no disk, and Hyperstack's closest flavor has 8 vCPU against 4.

Which is faster, Hyperstack or Replicate?

We have no comparable CPU measurement for either, so we cannot say. Both are built around Nvidia GPUs, and Hyperstack offers newer cards such as B200 and B300 that Replicate does not list.

Can I run Docker on Hyperstack or Replicate?

On Hyperstack, yes: VMs have root and its Ubuntu images come with or without Docker. On Replicate, Cog builds your model into a Docker image, but running Docker inside it is not documented. bochka sandboxes come with Docker, Docker Compose, systemd and /dev/kvm.

What do they cost when nothing is running?

A stopped Hyperstack VM bills at the full rate; a hibernated one bills only its disk and a kept IP, and restoring it needs the flavor in stock. Replicate deployments scale to zero, though private models bill a few idle minutes after traffic stops. On bochka a stopped sandbox costs nothing and keeps its disk.

How long can a job run?

A Replicate prediction times out after 30 minutes unless support raises it. Hyperstack documents no maximum runtime, so a VM runs until you stop, hibernate or delete it. bochka has no session cap on paid plans.

Are Hyperstack or Replicate good for AI coding agents?

Hyperstack can host one on a VM you set up yourself, and Replicate is built for serving models, not for agents that need a machine. bochka gives each agent a full VM with coding agents, a desktop and a credentials proxy built in, and your agent can still call either for GPU work.

Try bochka

Install the CLI, then run bochka new.

Product preview