Run Qwen2.5-Coder 32B
Ready in about 90 seconds. No install, no setup. From $0.89 per hour.
What it does
Qwen2.5-Coder 32B writes, explains, reviews, and fixes code across about ninety programming languages.
It is the closest open model to the commercial coding assistants, and it is the only realistic option if your employer will not let your source code leave the building.
You can point an editor at it. It speaks the standard API that most coding extensions expect, so it drops into an existing setup.
What you can ask it
Explain what this 4,000 line legacy file does.
You get backA structured walkthrough by section, with the risky parts called out.
Write the tests for this module.
You get backA test file covering the main paths and the edge cases it can infer.
Convert this Python service to Go, keeping the behavior.
You get backThe translated code plus a list of the places behavior could differ.
Review this change for security problems.
You get backSpecific findings with line references, not a generic checklist.
Your setup
We run this on an L40S. It writes at about 42 tokens per second and holds roughly 128,000 tokens of your codebase in context. $0.89 per hour, billed by the minute.
Why this card?
The model is about 19GB compressed to 4-bit, which fits a 24GB card. The reason we use a 48GB one is context. Coding work means pasting in large amounts of surrounding code, and a full context adds around 18GB. Running out of context is what causes the model to forget a file you showed it two messages ago, so the extra memory is the difference between a useful assistant and an annoying one.
- Memory
- 48GB
- Speed
- about 3 minutes per thousand lines
- Rate
- $0.89/hr
What will it cost you?
Billed by the minute on a L40S at $0.89 per hour. A batch job shuts the GPU off when the last thousand lines is done, so this is the whole cost.
Two ways to run it
Click and use it
A chat window, or connect your editor to the session endpoint.
Get early access Best for exploring.Hand us the whole pile
Run a review or a migration across a whole repository in one job.
Join the list Best for volume. The GPU shuts off automatically when it is done.Model facts
| Parameters | 32B |
|---|---|
| License | Apache 2.0 |
| Memory required | 24GB minimum (4-bit), 48GB recommended |
| Base model | Qwen2.5, further trained on code |
| Publisher | Alibaba Cloud |
| Recommended hardware | L40S, 48GB |
| Hugging Face | Qwen/Qwen2.5-Coder-32B-Instruct |
Want Qwen2.5-Coder 32B the day it opens?
We are testing the rental flow with a small group first. Leave an email and we will tell you when Qwen2.5-Coder 32B is ready to run. One message, no newsletter.
Qwen2.5-Coder 32B questions
Can I connect my editor to this?
Yes. The workspace exposes an OpenAI-compatible endpoint, which is what Continue, Cline, Zed, and most other coding extensions expect. Paste the session URL and the session key into your extension settings and it works. The URL stops working when the session ends.
Does my code get used for training?
No. Nothing that passes through your session is stored by us or sent anywhere else, and the disk is wiped when the session ends. That is the reason most people run a coding model here rather than through a commercial API.
Do I need to install anything to use Qwen2.5-Coder 32B?
No. We start a machine with Qwen2.5-Coder 32B already loaded and hand you a link. Everything runs in your browser, there is nothing to download, and nothing is left on your computer afterwards. It works the same on a Mac, a Windows laptop, or a Chromebook. A workspace is usually ready in about 90 seconds.
How much does it cost to run 20 thousand lines written?
About $0.94. Qwen2.5-Coder 32B takes about 3 minutes per thousand lines on the L40S we recommend, so 20 thousand lines written is roughly 63 minutes of GPU time at $0.89 per hour. Billing is by the minute, and a batch job shuts the GPU off the moment the last item finishes. The calculator above works this out for your own numbers.
Can I run Qwen2.5-Coder 32B on a cheaper card?
Sometimes, and the calculator will not always make it look worth it. The model is about 19GB compressed to 4-bit, which fits a 24GB card. The reason we use a 48GB one is context. Coding work means pasting in large amounts of surrounding code, and a full context adds around 18GB. Running out of context is what causes the model to forget a file you showed it two messages ago, so the extra memory is the difference between a useful assistant and an annoying one. If you want to try a different card anyway, the advanced catalog lets you pick one and shows the estimated time before you commit.
Can I use Qwen2.5-Coder 32B commercially?
Yes. Qwen2.5-Coder 32B is released under the Apache 2.0 license, which permits commercial use with no conditions worth worrying about. You can use the output in paid client work and in products you sell.
Own a GPU that sits idle? Put it to work. The average provider earns $180 to $420 per month per card.
List your GPU