bochka vs Lightning AI
Lightning AI is a cloud for building and training AI models: browser workspaces called Studios, GPUs, and a newer Sandbox API. bochka is built for AI agents: every sandbox is a full VM with Docker, a 60fps desktop, coding agents you drive from outside and public HTTPS ports.
Where bochka stands out
Coding agents you drive from outside
bochka ships Claude Code, Codex, pi, OpenCode and other agents inside each VM, and you drive the agents inside from outside with a prompt, steer and events API. Lightning's SDK runs commands, files and terminals, and its AI copilot is for ML work in a Studio. bochka's credentials proxy also keeps real AI keys out of the VM, where Lightning passes secrets in as environment variables.
A desktop and a browser in every VM
Each bochka VM has a 60fps desktop and Chromium, with computer use and browser control, and you can embed the stream in your own app. Lightning Studios open as VSCode or JupyterLab in the browser, or over SSH, with no desktop.
Public ports, raw TCP and your own IP
bochka puts token-protected HTTPS URLs on up to 50 ports of each VM, plus raw TCP and a dedicated IP. Lightning Studios share apps through a ports plugin, and its Sandbox API returns no external URL for a port at all.
The companies and products
bochka
bochka.xyz runs sandboxes for AI agents: each one is a full Linux VM with its own kernel.
You start a machine with one command or one API call, then SSH in, run Docker, open a desktop, or let a built-in agent (Claude Code, Codex, OpenCode and others) work in it. The disk stays between sessions, and a stopped sandbox costs nothing.
- Isolation
- Full VM per sandbox, own kernel
- Regions
- EU: Germany, Finland, France
Lightning AI
Lightning AI, the company behind PyTorch Lightning, runs a cloud for coding, training and deploying AI models, and in 2026 merged with the GPU cloud Voltage Park.
Its main product is AI Studio: a persistent cloud workspace with VSCode and JupyterLab in the browser, sudo, Docker and a button to switch between CPU and GPU machines. It also sells GPUs by the hour, model APIs and a Sandbox API that starts isolated Linux environments from Python or JavaScript. It is built for ML engineers and research teams, so this page compares it where it overlaps with bochka: cloud machines that agents and developers work in.
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. “3rd cheapest of 367” means 2 of the other 366 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 | bochka: price rank | Lightning AI: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month bochka fits fully, with Docker, root and a desktop on every VM; on Lightning you can build in a Studio, but there are no built-in coding agents and the Sandbox API gives no public URL for an app. | 3rd cheapest of 367 $634 a month | 221st of 367, with limits $2,752 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An always-on assistant on bochka keeps one small VM running; on Lightning only one free CPU Studio runs at a time, so every extra assistant is a paid 8 vCPU Studio. | 5th cheapest of 366 $657 a month | 308th of 366, with limits $6,956 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month | 186th of 345, with limits $2,000 a month | 295th of 345, with limits $721 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Sandboxes for your users need many small machines with public URLs; Lightning's paid CPU Studios start at 8 vCPU / 32 GiB and its Sandbox API has no published price. | Cheapest of 361 $209 a month | 290th of 361, with limits $2,264 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 34th of 360, with limits $2,000 a month | 287th of 360, with limits $10,939 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 10th cheapest of 364 $200 a month | 277th of 364, with limits $885 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month | 4th cheapest of 377 $86 a month | 273rd of 377, with limits $408 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Both fit: Lightning's one free Studio and browser VSCode make a good personal box, and bochka keeps a 4 vCPU VM on all month with a desktop. | 18th cheapest of 368 $26 a month | 109th cheapest of 368 $324 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month bochka streams a 60fps desktop and Chromium for computer use; Lightning has no desktop, only browser IDEs and SSH. | 6th cheapest of 369 $200 a month | 218th of 369, with limits $797 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month bochka ships Chromium with browser control in each VM; on Lightning you install and run a headless browser yourself. | 7th cheapest of 367 $200 a month | 287th of 367, with limits $792 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 6th cheapest of 346 $90 a month | 313th of 346, with limits $1,073 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 71st of 338, with limits $1,800 a month | 288th of 338, with limits $21,450 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 2nd cheapest of 360 $200 a month | 279th of 360, with limits $526 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month Only Lightning fits: it rents GPUs from T4 to B200 by the second, and bochka has no GPUs. | Can't run it | 72nd of 98, with limits $1,117 a month |
| 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 |
Key features
Every feature Battleships tracks (43)
| Feature | bochka | Lightning AI |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | Not published |
| Root | Yes | Yes |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | No |
| Browser | Yes | No |
| Computer-use API | Yes | No |
| SSH | Yes | Yes |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | Yes |
| Fork/clone | Yes | Yes |
| Live fork (no pause) | Yes | Yes |
| Memory fork | No | No |
| Custom image (Docker or snapshot) | Yes | Yes |
| ≥24 h sessions | Yes | Yes |
| Idle auto-stop | No | Yes |
| Wake on request | No | Yes |
| HTTPS preview URLs | Yes | Yes |
| Raw TCP inbound | Yes | Not published |
| Public IPv4 | No | No |
| Egress allowlist | No | Yes |
| Secret proxy | Yes | No |
| Secret proxy for any API | Yes | No |
| Preinstalled agents | Yes | No |
| Agent harness API | Yes | No |
| MCP server | No | No |
| GPU | No | Yes |
| Windows | No | No |
| Arm64 | No | No |
| Self-hosting | No | Yes |
| BYOC | No | Yes |
| Open source | No | No |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Yes |
| Usage per sandbox (API) | Yes | No |
| Audit logs | No | Yes |
| SSO | No | Yes |
| SOC 2 | No | Yes |
| HIPAA | No | Yes |
| EU data residency | Yes | Yes |
| Zero data retention | Yes | Not published |
Performance
There is no comparable CPU measurement for Lightning AI: it was not in the public benchmark where bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB. ComputeSDK has measured how fast Lightning's Sandbox API starts, which is a different test, so we do not compare CPU speed here.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Lightning AI has 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 | bochka, a month | Lightning AI, a month |
|---|---|---|
| One coding agent, workdays 4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days | $20 · $20/mo | $50 · Pro (monthly) |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $50 · Pro (monthly) |
| 50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside | $317 · $20/mo × 3 seats | $4,089 · Free · with limits |
| Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once | $200 · $100/mo × 2 seats | $2,295 · Free · with limits |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $2,000 · $2000/mo · with limits | $5,072 · Free · with limits |
bochka: You pick a monthly plan from $20, and all of it becomes usage credit spent per second on running sandboxes. A stopped sandbox costs nothing and keeps its disk.
Lightning AI: Lightning bills per second from prepaid credits, and Pro costs $50 a month with 40 credits included. One Default CPU Studio (4 cores / 16 GB) runs free; every other machine is paid, and the Sandbox API has no published price, so the scenarios use Studio prices. Storage above 10 GB is billed, and egress prices are not published.
Limits in those scenarios
Lightning lets one free CPU Studio run at a time and publishes no price for a second 4-core Studio, so the scenarios put every extra machine on the 8 vCPU / 32 GiB Large shape. That is why 50 agents in parallel come to $4,089 a month and 300 small sandboxes for your users are priced on machines with four times the vCPU they need. Lightning does not publish how many CPU machines can run at once, so the fleet scenarios are marked as not confirmed. On bochka the limit is a sandbox count (100 to 2,000 by plan) plus starts per minute, so 1,000 short jobs at once get queued and need more starts per day than the $2,000 plan allows.
| Scenario | bochka | Lightning AI |
|---|---|---|
| One coding agent, workdays | None that bite |
|
| One agent that never stops | None that bite |
|
| 50 agents in parallel | None that bite |
|
| Sandboxes for your users | None that bite |
|
| 1,000 short jobs at once |
|
|
Which one should you use?
Use Lightning AI when
your work needs GPUs, a full ML workflow with notebooks, training and model hosting, or regions and private clouds outside the EU.
- GPUs from T4 to B200. Useful if your agents train, fine-tune or test models: Lightning rents Nvidia GPUs by the second, from T4 to B200.
- Built for ML teams. Useful for ML teams: Studios come with VSCode, JupyterLab, shared drives, training jobs and model hosting in one place.
- More regions, your own cloud and egress rules. Lightning runs in US, EU and Asia regions, in your own VPC or on-prem, and sandboxes can be limited to chosen IP ranges.
Use bochka when
you run AI agents: coding agents you drive from outside, a desktop and browser for computer use, public ports for the apps they build, on a full VM with Docker and root.
- Coding agents you drive from outside. On Lightning you install each agent, write the code that starts, prompts and watches it, and put real keys in its environment, where code the agent runs can read them.
- A desktop and a browser in every VM. An agent that has to click through a web app or test a UI needs a screen. On Lightning you set up a virtual display, a browser and a way to watch it yourself.
- Public ports, raw TCP and your own IP. When an agent builds an app, you or your users need to open it. On Lightning's Sandbox API the port is reachable only from inside its network, so you add your own ingress.
Questions people ask
Is bochka cheaper than Lightning AI?
In the scenarios we priced, yes: one coding agent on workdays comes to $20 a month on bochka and $50 on Lightning, and one agent running all month comes to $26 versus $50. 50 agents in parallel come to $317 versus $4,089, and sandboxes for your users $200 versus $2,295. Lightning's Sandbox API has no published price, so these use its Studio prices.
Is Lightning AI a sandbox for AI agents?
Partly. Lightning is mostly a platform for building and training models, with Studios, GPUs and model hosting. It added a Sandbox API that starts isolated Linux environments with sudo and snapshots, but it has no desktop, no built-in coding agents and no public URL for sandbox ports.
Does bochka have GPUs like Lightning AI?
No. bochka runs Linux VMs on dedicated AMD Ryzen 9 9950X (Zen 5) hosts and has no GPUs. If your agents need to train or run models on a GPU, Lightning rents GPUs from T4 to B200.
Can I run Docker in both?
Yes. Lightning Studios run Docker like a laptop and sandboxes can start from any OCI image. bochka VMs have Docker, root, systemd and /dev/kvm on every machine, so emulators and nested VMs work too.
Where does my data run?
bochka runs only in the EU (Germany, Finland and France), with EU data residency and a zero data retention option. Lightning stores data in US, EU or Asia regions when you link a cloud account, and Enterprise can run in your own VPC or on-prem.
Is bochka SOC 2 compliant?
SOC 2 is in progress. We're open about where we stand: live status at status.bochka.xyz and our data retention policy at docs.bochka.xyz/data-retention. Lightning lists SOC 2 (type 2) and HIPAA on its Enterprise plan.
Start a sandbox
Install the CLI, then run bochka new.
Product preview