# Run Qwen2.5 72B

The strongest open model for reasoning, math, and non-English work. Ready in about 90 seconds on an A100 80GB, from $1.35/hr. No install, no code.

## What it does

Qwen2.5 72B is the open model that holds up best on problems with a right answer: math, logic, structured data, and multi-step instructions.

It is also the strongest choice if your work is not in English. It handles Chinese, Spanish, French, Arabic, Japanese, and about two dozen other languages at close to its English quality.

It costs the same to run as Llama 3.3 70B. Pick Qwen when the task has a correct answer, and Llama when the task is writing.

## Your setup

We run Qwen2.5 72B on an A100 80GB with 80GB of memory, at $1.35 per hour, billed by the minute. It takes about 36 seconds per thousand words.

Qwen2.5 72B handles a context of up to 128,000 tokens, which is roughly a 300 page document. That context is what eats memory. The model itself is about 41GB compressed, and a full-length context can add another 20GB on top. An 80GB card covers both. A 48GB card runs the model but caps you at short documents, which removes the main reason to use it.

## What it costs

100 thousand words written takes about 60 minutes of GPU time, which is about $1.35.

## Two ways to run it

- **Open the app.** A chat window with document upload, in your browser.
- **Run a batch job.** Row-by-row processing of a spreadsheet, with the GPU shutting off at the end.

## Model facts

| Field | Value |
|---|---|
| Parameters | 72B |
| License | Qwen License |
| Memory required | 48GB minimum (4-bit), 80GB recommended |
| Base model | Trained from scratch by Alibaba Cloud |
| Publisher | Alibaba Cloud |
| Recommended hardware | A100 80GB, 80GB |
| Category | [Language models](https://gpuvault.io/working-on/language-models/) |

## Questions people ask

### Do I need to install anything to use Qwen2.5 72B?

No. We start a machine with Qwen2.5 72B 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 100 thousand words written?

About $1.35. Qwen2.5 72B takes about 36 seconds per thousand words on the A100 80GB we recommend, so 100 thousand words written is roughly 60 minutes of GPU time at $1.35 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 72B on a cheaper card?

Sometimes, and the calculator will not always make it look worth it. Qwen2.5 72B handles a context of up to 128,000 tokens, which is roughly a 300 page document. That context is what eats memory. The model itself is about 41GB compressed, and a full-length context can add another 20GB on top. An 80GB card covers both. A 48GB card runs the model but caps you at short documents, which removes the main reason to use it. If you want to try a different card anyway, the advanced catalog lets you pick one and shows the estimated time before you commit.

### What are the license restrictions on Qwen2.5 72B?

Qwen2.5 72B is released under the Qwen License. That is not a permissive license, so read it before you use the output in paid work. We show the license on every model page precisely because this catches people out, and there is usually a permissively licensed alternative in the same category.

### What happens to my files after I stop?

Anything you save into your workspace folder stays in your account and is there when you come back. Everything else is destroyed: the machine is terminated and its disk is wiped. Batch job inputs and outputs are deleted from our storage 24 hours after you download them. We do not read your files or use them to train anything.

### What if it fails partway through?

You are not charged for work that did not complete, and failed sessions are refunded automatically. Batch jobs write each output as it finishes rather than at the end, so a failure at item 380 of 400 leaves you 379 usable files and a message saying what broke. Restarting picks up where it stopped rather than redoing the work.

---

*Source: https://gpuvault.io/model/qwen-2-5-72b/ — GPUVault, GPU rental by the minute.*