Coding assistants

Connect your editor to a coding model running on a machine only you can reach. Nothing you write leaves the session.

SOC 2 controls 142 verified providers 99.2% uptime Ready in about 90 seconds

What do you need to do?

Editor autocomplete Fast enough that you do not wait for it. Explain legacy code A walkthrough of the file nobody understands. Write tests Coverage for a module you already have. Review a change Specific findings with line references. Migrate a codebase One language to another, across a repository. Fill in the middle Completion with code above and below the gap.

Models for this work

All categories
Best quality Qwen2.5-Coder 32B The strongest open coding model. Runs privately on your own code. from $0.89/hr about 3 minutes per thousand lines Fastest DeepSeek-Coder V2 Lite A fast, cheap coding model. Good for autocomplete and small fixes. from $0.45/hr about 2 minutes per thousand lines Codestral 22B Built specifically for filling in code in the middle of a file. from $0.45/hr about 2 minutes per thousand lines

Which one should I pick?

ModelBest forQualitySpeedPrice
Qwen2.5-Coder 32B Serious coding work on code that cannot leave your network Excellent Fast $0.89/hr
DeepSeek-Coder V2 Lite Editor autocomplete and everyday small changes Very good Very fast $0.45/hr
Codestral 22B Editor completion, on work that is not commercial Very good Fast $0.45/hr

Prices are the hourly rate for the hardware we recommend for each model, billed by the minute. Every model page has a calculator that turns that into a total for your job.

Two people working together at a laptop
NO SURPRISES

You will never get a surprise bill.

  • A hard spend cap. New accounts start at $25. Nothing starts that would go past it.
  • Auto-shutoff when idle. Sessions stop after 15 minutes, with a warning at 12.
  • A live cost meter. Spent so far and budget left, on screen the whole time.
COMING SOON

Renting opens shortly. Want to know when?

We are testing the rental flow with a small group before we open it to everyone. Leave an email and we will tell you the day it opens. One message, no newsletter.

Questions people ask

Can I connect my existing editor?

Yes. Each session exposes an OpenAI-compatible endpoint, which is the interface Continue, Cline, Zed, Aider, and most other coding extensions already speak. Paste the session URL and key into your extension settings and it behaves like any other provider. The URL stops working when the session ends, so you paste a new one next time.

Does my source code get stored or used for training?

No. Nothing that passes through your session is retained by us or sent anywhere else, and the machine and its disk are destroyed when the session ends. This is the reason most people run a coding model here rather than through a commercial API, and it is usually what makes the difference to a security team.

Are open coding models good enough to be worth it?

Qwen2.5-Coder 32B is close to the commercial assistants on everyday work: writing functions, explaining code, generating tests, and reviewing changes. It is behind on very large multi-file refactors and on the newest frameworks. If your constraint is that code cannot leave your network, it is comfortably good enough to be worth using. If you have no such constraint, evaluate it against what you have rather than assuming either way.

How much does it cost to run a coding assistant all day?

Between $3.60 and $7.10 for an eight hour day, depending on the model. DeepSeek-Coder V2 Lite on an RTX 4090 is $0.45 per hour and Qwen2.5-Coder 32B on an L40S is $0.89. Idle auto-shutoff means you are not billed while you are in a meeting: the session stops after 15 minutes of no activity and restarts in about 90 seconds.

Can I use these models at work?

Qwen2.5-Coder 32B is Apache 2.0 and DeepSeek-Coder V2 Lite permits commercial use, so both are fine for employed or client work. Codestral is not: its license covers personal projects, research, and evaluation only, and excludes commercial and internal business use. That is a real restriction rather than a technicality, and it is why the license appears on every model page.

What happens to my session when I close my laptop?

It keeps running until it hits the idle timeout, which defaults to 15 minutes of no activity, then shuts itself down. You get a warning at 12 minutes with a button to keep it running. Anything you saved into your workspace folder survives; the machine does not. Maximum session length defaults to 4 hours, and you can raise both limits deliberately in your account.

Own a GPU that sits idle? Put it to work. The average provider earns $180 to $420 per month per card.

List your GPU