WEBVTT 1 00:00:00.570 --> 00:00:04.620 So where do you find those open models? 2 00:00:04.620 --> 00:00:08.310 And with that, I of course mean the ones I just described, 3 00:00:08.310 --> 00:00:10.560 but maybe also others 4 00:00:10.560 --> 00:00:14.430 because it's of course not just these four providers 5 00:00:14.430 --> 00:00:17.970 that do publish open Large Language Models. 6 00:00:17.970 --> 00:00:20.280 Well, of course if you do want 7 00:00:20.280 --> 00:00:22.710 to learn more about Meta's Llama models 8 00:00:22.710 --> 00:00:24.720 or Google's Gemma models, 9 00:00:24.720 --> 00:00:28.950 their official websites are a great place to go. 10 00:00:28.950 --> 00:00:32.970 But to get a general overview of available models 11 00:00:32.970 --> 00:00:35.910 and to explore available models 12 00:00:35.910 --> 00:00:37.920 and also find new models, 13 00:00:37.920 --> 00:00:41.010 which you maybe haven't heard about before, 14 00:00:41.010 --> 00:00:45.361 Hugging Face is the defacto standard place to go. 15 00:00:45.361 --> 00:00:49.440 Huggingface.co is a website 16 00:00:49.440 --> 00:00:53.790 that's all about open AI models in the end. 17 00:00:53.790 --> 00:00:55.770 They have a models catalog 18 00:00:55.770 --> 00:00:58.350 where all the big players like Meta, Google, 19 00:00:58.350 --> 00:01:02.370 but also everybody else that's creating AI models, 20 00:01:02.370 --> 00:01:06.900 publishes those models or lists those models. 21 00:01:06.900 --> 00:01:11.900 So in that model catalog, you'll find a plethora 22 00:01:11.970 --> 00:01:16.970 of open Large Language and also other AI models. 23 00:01:17.280 --> 00:01:18.750 You will, for example, 24 00:01:18.750 --> 00:01:22.380 find information about Google's Gemma models in there, 25 00:01:22.380 --> 00:01:24.060 about the Qwen model, 26 00:01:24.060 --> 00:01:27.030 which is another popular open model, 27 00:01:27.030 --> 00:01:29.759 about Deepseek models, 28 00:01:29.759 --> 00:01:32.580 Microsoft's open models, 29 00:01:32.580 --> 00:01:37.080 and of course also Meta's Llama models, for example. 30 00:01:37.080 --> 00:01:40.170 And again, many, many more. 31 00:01:40.170 --> 00:01:43.650 Now in this model catalog, you can sort by trending 32 00:01:43.650 --> 00:01:47.130 or by most downloads, to also play around 33 00:01:47.130 --> 00:01:51.210 with the models most other people's maybe using. 34 00:01:51.210 --> 00:01:53.640 You can also sort based on recency 35 00:01:53.640 --> 00:01:57.420 to find out which new models have been published 36 00:01:57.420 --> 00:02:00.570 or what the most liked models are. 37 00:02:00.570 --> 00:02:01.920 It's also worth noting 38 00:02:01.920 --> 00:02:03.780 that Hugging Face is not just 39 00:02:03.780 --> 00:02:06.900 about Large Language Models, though. 40 00:02:06.900 --> 00:02:09.060 You can filter for which kind 41 00:02:09.060 --> 00:02:12.000 of AI model you are interested in. 42 00:02:12.000 --> 00:02:16.200 You, for example, can also explore computer vision models 43 00:02:16.200 --> 00:02:18.660 that are able to analyze images 44 00:02:18.660 --> 00:02:21.750 or models that are able to generate images 45 00:02:21.750 --> 00:02:25.080 or models that can handle audio. 46 00:02:25.080 --> 00:02:26.760 So there's a broad variety 47 00:02:26.760 --> 00:02:29.070 of models listed here in this course, 48 00:02:29.070 --> 00:02:31.860 however, will focus on the Large Language Models, 49 00:02:31.860 --> 00:02:32.952 which are arguably 50 00:02:32.952 --> 00:02:37.143 the most popular open models being used out there. 51 00:02:38.430 --> 00:02:40.800 And what's really great about Hugging Face is 52 00:02:40.800 --> 00:02:44.190 that you don't just get this model catalog 53 00:02:44.190 --> 00:02:47.190 where you have this list of models, 54 00:02:47.190 --> 00:02:50.940 but of course you can also inspect those different models 55 00:02:50.940 --> 00:02:52.020 and learn more about them. 56 00:02:52.020 --> 00:02:55.860 For example, if I inspect the Google Gemma 3, 57 00:02:55.860 --> 00:02:59.880 27 billion parameters model here, 58 00:02:59.880 --> 00:03:02.250 which is just one possible model 59 00:03:02.250 --> 00:03:04.560 you could run locally as well, 60 00:03:04.560 --> 00:03:07.530 you get such a model card 61 00:03:07.530 --> 00:03:11.910 and every model hosted on Hugging Face has such a card. 62 00:03:11.910 --> 00:03:14.730 And this card simply describes the model. 63 00:03:14.730 --> 00:03:18.360 It describes the capabilities of this model, for example, 64 00:03:18.360 --> 00:03:20.910 that the Gemma model is multimodal. 65 00:03:20.910 --> 00:03:24.330 It can handle text and image input. 66 00:03:24.330 --> 00:03:27.270 It describes the context window and anything else 67 00:03:27.270 --> 00:03:30.360 the provider of that model, Google, in this case, 68 00:03:30.360 --> 00:03:32.433 felt worth sharing. 69 00:03:33.540 --> 00:03:36.360 So it's a good way to understand what a model can do. 70 00:03:36.360 --> 00:03:38.340 You'll also find some code examples 71 00:03:38.340 --> 00:03:41.220 on how you could use this model in your code, 72 00:03:41.220 --> 00:03:44.340 though I'll get back to using these models locally, 73 00:03:44.340 --> 00:03:46.740 including programmatically a little bit later 74 00:03:46.740 --> 00:03:47.613 in this course. 75 00:03:48.960 --> 00:03:53.430 And you'll also learn more about the size of this model, 76 00:03:53.430 --> 00:03:56.160 so the number of parameters it has. 77 00:03:56.160 --> 00:03:59.130 So that would be 27 billion parameters here 78 00:03:59.130 --> 00:04:02.940 for this Gemma 3 27 billion model. 79 00:04:02.940 --> 00:04:05.130 The data type of these parameters 80 00:04:05.130 --> 00:04:08.010 and some other things to which I'll get back later, 81 00:04:08.010 --> 00:04:10.680 which are important to understand when it comes 82 00:04:10.680 --> 00:04:14.160 to running these models locally on your system. 83 00:04:14.160 --> 00:04:17.070 But again, I will get back to that later. 84 00:04:17.070 --> 00:04:18.600 The key takeaway here is 85 00:04:18.600 --> 00:04:23.190 that this Hugging Face models catalog is the place 86 00:04:23.190 --> 00:04:27.265 to go to explore all available open models 87 00:04:27.265 --> 00:04:32.040 you could consider running locally or on your own servers. 88 00:04:32.040 --> 00:04:35.310 It's there for all as a great place to explore 89 00:04:35.310 --> 00:04:39.180 not just the big well-known model providers 90 00:04:39.180 --> 00:04:42.000 and their models, but also more niche 91 00:04:42.000 --> 00:04:45.843 or upcoming open models that might be worth exploring.