# Run Mistral Small 3.1

A quick, cheap assistant that also reads images. Good default for volume work. Ready in about 90 seconds on an RTX 4090, from $0.45/hr. No install, no code.

## What it does

Mistral Small 3.1 is a 24 billion parameter model that answers about as well as models three times its size on everyday tasks, and does it much faster.

It reads images as well as text, so you can hand it a screenshot, a scanned form, or a photograph of a receipt and ask questions about it.

It is the right pick when you have thousands of small jobs rather than a few hard ones: classifying tickets, tagging products, summarizing reviews, pulling fields out of forms.

## Your setup

We run Mistral Small 3.1 on an RTX 4090 with 24GB of memory, at $0.45 per hour, billed by the minute. It takes about 24 seconds per thousand words.

At 24 billion parameters the 4-bit version is roughly 14GB, so it fits on a 24GB card with about 10GB left for the conversation and any images you send it. Moving up to a data center card would triple the hourly cost and gain very little, because this model is small enough that the RTX 4090 is not the bottleneck.

## What it costs

200 thousand words written takes about 80 minutes of GPU time, which is about $0.60.

## Two ways to run it

- **Open the app.** A chat window in your browser. Drag images straight into the conversation.
- **Run a batch job.** The mode most people want here. Upload a spreadsheet, describe the task once, and collect the finished file.

## Model facts

| Field | Value |
|---|---|
| Parameters | 24B |
| License | Apache 2.0 |
| Memory required | 16GB minimum (4-bit), 24GB recommended |
| Base model | Trained from scratch by Mistral AI |
| Publisher | Mistral AI |
| Recommended hardware | RTX 4090, 24GB |
| Category | [Language models](https://gpuvault.io/working-on/language-models/) |

## Questions people ask

### Can I use Mistral Small 3.1 commercially?

Yes, without conditions. It is released under Apache 2.0, which is the most permissive license in common use. You can use it in a product you sell, modify it, and you do not have to credit anyone or publish your changes.

### Do I need to install anything to use Mistral Small 3.1?

No. We start a machine with Mistral Small 3.1 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 200 thousand words written?

About $0.60. Mistral Small 3.1 takes about 24 seconds per thousand words on the RTX 4090 we recommend, so 200 thousand words written is roughly 80 minutes of GPU time at $0.45 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 Mistral Small 3.1 on a cheaper card?

Sometimes, and the calculator will not always make it look worth it. At 24 billion parameters the 4-bit version is roughly 14GB, so it fits on a 24GB card with about 10GB left for the conversation and any images you send it. Moving up to a data center card would triple the hourly cost and gain very little, because this model is small enough that the RTX 4090 is not the bottleneck. 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 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/mistral-small-3-1/ — GPUVault, GPU rental by the minute.*