---
title: 'Setting up DeepSeek R1 on Vast.ai'
source: 'https://youtube.com/watch?v=542xENIxKFU'
video_id: '542xENIxKFU'
date: 2026-07-28
duration_sec: 358
---

# Setting up DeepSeek R1 on Vast.ai

> Source: [Setting up DeepSeek R1 on Vast.ai](https://youtube.com/watch?v=542xENIxKFU)

## Summary

This video demonstrates how to set up DeepSeek R1 (70B distilled) on a 2x RTX 4090 machine using Vast.ai. The process involves selecting a template, renting a GPU instance, and using Ollama with Open WebUI to run the model.

### Key Points

- **Introduction** [0:06] — Travis from Vast.ai shows how to spin up DeepSeek R1 70B on a 2x RTX 4090 machine.
- **Software Stack** [0:54] — Uses Ollama to run the model and Open WebUI for the chat interface.
- **Template Selection** [1:25] — Selects the 'ollama webui' template and allocates 60 GB storage.
- **GPU Selection** [1:40] — Filters for 2x RTX 4090 machines and selects one in Iceland.
- **Instance Launch** [2:21] — Instance loads in about a minute; opens instance portal showing running processes.
- **Open WebUI Setup** [2:55] — Creates admin account and navigates to settings to add a model.
- **Model Download** [3:55] — Enters model tag 'deepseek-r1:70b' and downloads it.
- **Model Running** [4:49] — DeepSeek R1 70B is loaded and ready for chat; shows thinking step.
- **Cost and Connectivity** [5:10] — Costs about $0.80/hour; also supports API, Jupyter, and SSH.

### Conclusion

DeepSeek R1 70B can be easily deployed on Vast.ai using a 2x RTX 4090 machine with Ollama and Open WebUI, costing around $0.80 per hour.

## Transcript

hello hey this is Travis from Bast I
wanted to make a quick video to show you
how to spin up deep seeks R1 uh that's
distilled down to the 70 billion uh size
it fits very well on a 2X 490 machine so
I just wanted to show you really quick
how to spin that up on vast so I'll go
ahead and jump into it so we have uh our
console interface here and if you're new
to vast and just getting started we have
a quick start guide that goes over some
of the basics of how to create an
account add credit um set up your
template and select a template filter
machines and then R to
GPU we'll go ahead and jump into uh the
guide that we have for this use
case
and to run this model uh the Deep seek
r170 billion we're going to use AMA and
web UI AMA is the software stack that's
going to run it and web UI actually set
up the UI that you can access um to
prompt the model very similar to what
you do with uh open AI so let's go ahead
and get started I have credit and my
account is set up so I'm going to go
ahead and pick the correct
template which is this uhan web UI
I'm going to allocate enough storage uh
for this model I only need about 40 gab
but I'll just kind of leave it around 60
and then I'm going to want a
2X
4090 so I'll go ahead and put that in
here now I'm also going to select um I
don't know how long uh you plan on using
the model but I'm only going to use this
for a couple days so I'll just kind of
move the max instance duration down not
going to rent this for months um so
here's all the uh availability of
machines uh the ones that are that have
the blue label are in a trusted Data
Center and um I'll go ahead and pick uh
this 2x 490 in Iceland now I'm going to
click Rent Now what's going to happen is
the instance going to spin up and it's
going to start loading this
template all right looks like a minute
has passed and now the instance is uh
loaded so I'm going to go ahead and
click this open Button
and this is going to open our instance
portal uh which basically is uh shows
all the process that that are running on
this instance right now and you can see
uh the open web UI is running on this
port and the API is running on this
boort so I'll go ahead and click
this and this will start up and launch
the web
UI this is a open source software to run
and and chat with
llms so when you first boot this up you
have to make a
name uh
email and a password for the
admin account and then it's going to
show you kind of what's new in open web
UI and uh it's going to load you into a
screen that has no model selected so
you're going to have to download a
model I think there's a few ways to add
a model um in this interface I kind of
just explore but if you go into the
admin panel and then click
settings
models and then you
can add a
model and uh this model that I'm running
is from ama.com so you just need to put
in the
tag and the tag is
is just the name of the model which is
right
here so I'm going to grab the tag I'm
going to drop it in
here hit the download button and now
it's going to start downloading the
model it'll take a few minutes to
download to go get a cup of
coffee but once it gets to 100% now the
Deep seek model is um
load it onto this
instance and you can simply start a new
chat and select the model after it
downloads it also could take a little
bit of additional time to load
it and there it is this is the Deep seek
r170 billion model running on a 2X
490
and there we can see
it's a reasoning model so it has a
thinking step uh that it's going to give
the
answer can see on vast um I'm paying
about 80 cents an hour right now for
this if I wanted to keep this machine
longer there's a reserved instance
discount um but I'm uh just going to
keep it for uh a few hours to to play
around with this a bit and if I wanted
to run the API the API uh is available
on uh this
port and uh Jupiter is also available
along with SSH so there's a lot of
different ways you can connect to this
machine um but it's always very simple
to just use this sort of chat interface
that loads onto it well that's deep seek
r170 billion on vast AI thanks