WEBVTT

1
00:00:00.270 --> 00:00:02.010
<v Maximilian>So how do you use Ollama</v>

2
00:00:02.010 --> 00:00:05.370
to run open models on your system?

3
00:00:05.370 --> 00:00:07.740
Well, if you run the ollama command

4
00:00:07.740 --> 00:00:10.440
in your terminal, in your command line,

5
00:00:10.440 --> 00:00:13.350
you'll see a bunch of available commands

6
00:00:13.350 --> 00:00:17.220
that you can combine with the ollama command.

7
00:00:17.220 --> 00:00:20.250
For example, the very important run command,

8
00:00:20.250 --> 00:00:24.960
which can be used to run a model as we learn here,

9
00:00:24.960 --> 00:00:27.870
and that's indeed the command we need to execute,

10
00:00:27.870 --> 00:00:32.870
well, to run a Large Language Model on our system here.

11
00:00:32.910 --> 00:00:36.780
So ollama run, but then what?

12
00:00:36.780 --> 00:00:40.920
Well, then you need to specify the model identifier

13
00:00:40.920 --> 00:00:43.920
of the model that you wanna run here,

14
00:00:43.920 --> 00:00:46.380
and you'll find that in the model's catalog

15
00:00:46.380 --> 00:00:50.130
on the Ollama website, so on Ollama.com.

16
00:00:50.130 --> 00:00:54.690
Now, there also are other ways of creating models

17
00:00:54.690 --> 00:00:56.400
or of running models locally,

18
00:00:56.400 --> 00:00:57.930
but I'll get back to those later.

19
00:00:57.930 --> 00:01:00.900
Using the model's catalog is the most straightforward

20
00:01:00.900 --> 00:01:05.043
and standard way of running models with Ollama.

21
00:01:06.180 --> 00:01:10.770
So if you go here, you'll see a list of supported AI models,

22
00:01:10.770 --> 00:01:13.440
and it's a very, very long list

23
00:01:13.440 --> 00:01:16.650
because Ollama essentially supports

24
00:01:16.650 --> 00:01:21.300
all the popular and important open Large Language Models

25
00:01:21.300 --> 00:01:23.850
you also find on Hugging Face.

26
00:01:23.850 --> 00:01:28.350
And indeed, the Ollama team typically is pretty fast

27
00:01:28.350 --> 00:01:32.070
when it comes to adding support for a new model.

28
00:01:32.070 --> 00:01:36.090
So therefore, in this Ollama model's catalog here,

29
00:01:36.090 --> 00:01:39.120
you typically find all the important, latest,

30
00:01:39.120 --> 00:01:42.270
and most popular open Large Language Models

31
00:01:42.270 --> 00:01:45.150
you can run with Ollama.

32
00:01:45.150 --> 00:01:48.270
For example, here, you also find deepseek-r1,

33
00:01:48.270 --> 00:01:50.970
including the big one.

34
00:01:50.970 --> 00:01:53.550
I'll get back to these different versions here in a second,

35
00:01:53.550 --> 00:01:57.240
but you could in theory run the large one,

36
00:01:57.240 --> 00:01:59.460
the one that shook up the AI industry

37
00:01:59.460 --> 00:02:02.640
in early 2025 with Ollama.

38
00:02:02.640 --> 00:02:05.910
Though I will say that even with quantization,

39
00:02:05.910 --> 00:02:09.810
you probably don't have the hardware to run that.

40
00:02:09.810 --> 00:02:12.270
But all these smaller models,

41
00:02:12.270 --> 00:02:15.540
especially up to the 32 billion parameters,

42
00:02:15.540 --> 00:02:17.220
would definitely be doable,

43
00:02:17.220 --> 00:02:20.163
depending on the kind of hardware you have, of course.

44
00:02:21.090 --> 00:02:25.350
And then there are plenty of other popular models as well.

45
00:02:25.350 --> 00:02:28.470
For example, mistral models, llama models.

46
00:02:28.470 --> 00:02:31.770
The gemma3, and also the older gemma models,

47
00:02:31.770 --> 00:02:33.150
which are really amazing

48
00:02:33.150 --> 00:02:35.640
and which do have vision capabilities,

49
00:02:35.640 --> 00:02:39.510
so which can also be used to analyze images, for example.

50
00:02:39.510 --> 00:02:40.710
And you can, in general, of course,

51
00:02:40.710 --> 00:02:42.870
simply browse that catalog to find out

52
00:02:42.870 --> 00:02:46.170
which model sounds most interesting to you.

53
00:02:46.170 --> 00:02:49.680
As mentioned, you can also run models in another way,

54
00:02:49.680 --> 00:02:51.540
and I'll get back to that later.

55
00:02:51.540 --> 00:02:54.093
But this official model catalog is a good start.

