1
00:00:00,000 --> 00:00:06,000
So guys, already in our previous video we have discussed about if you probably see over here, we have

2
00:00:06,000 --> 00:00:08,000
discussed about this entire evolution of large language model.

3
00:00:09,000 --> 00:00:13,000
I went through this amazing website which is called as Artificial Analysis dot AI.

4
00:00:13,000 --> 00:00:19,000
And here, uh, specifically comparison is done with respect to various models that are probably there

5
00:00:19,000 --> 00:00:20,000
right now.

6
00:00:20,000 --> 00:00:26,000
You know, so here you can actually see with respect to the quality right now, GPT four is having the

7
00:00:26,000 --> 00:00:27,000
best quality, right.

8
00:00:27,000 --> 00:00:34,000
If you see cloudy 3.5 sonnet, then you have Gemini 1.5 Pro, then you have cloudy three opus, then

9
00:00:34,000 --> 00:00:36,000
you have Gemini 1.5 flash.

10
00:00:36,000 --> 00:00:41,000
Then you have llama three model which is 70 billion parameters, which is again an open source.

11
00:00:41,000 --> 00:00:42,000
Right.

12
00:00:42,000 --> 00:00:45,000
So here the combination of both open source and paid model has been probably considered.

13
00:00:45,000 --> 00:00:48,000
You also have llama three which is an 8 billion parameter.

14
00:00:48,000 --> 00:00:52,000
Then you have this Mistral eight into seven b then GPT 53.5 turbo.

15
00:00:52,000 --> 00:00:52,000
Right.

16
00:00:52,000 --> 00:00:58,000
So with respect to quality right now, GPT four is basically having the best.

17
00:00:58,000 --> 00:01:04,000
If you see with respect to speed right now, Gemini 1.5 flash is having the best speed with respect

18
00:01:04,000 --> 00:01:05,000
to inferencing.

19
00:01:05,000 --> 00:01:07,000
Then you have this llama three.

20
00:01:08,000 --> 00:01:09,000
Then you have cloudy three haiku.

21
00:01:10,000 --> 00:01:15,000
Um, with respect to speed basically means uh, how well the optimized the specific models are.

22
00:01:15,000 --> 00:01:20,000
And you can probably go ahead and check this with respect to price USD per 1 million token.

23
00:01:20,000 --> 00:01:23,000
And right now you can see cloudy three opus is the highest.

24
00:01:23,000 --> 00:01:25,000
The best is llama three.

25
00:01:25,000 --> 00:01:30,000
And this is with respect to again inferencing if it is specifically deployed in a cloud right.

26
00:01:30,000 --> 00:01:33,000
So definitely go ahead and check out this particular website.

27
00:01:33,000 --> 00:01:39,000
Here you will be able to get many more information like which all models are basically best in general

28
00:01:39,000 --> 00:01:39,000
ability.

29
00:01:39,000 --> 00:01:40,000
Right.

30
00:01:40,000 --> 00:01:46,000
So here from the chatbot arena this data is taken to GPT four is high reasoning and knowledge over here.

31
00:01:46,000 --> 00:01:48,000
ML GPT four is the best.

32
00:01:48,000 --> 00:01:51,000
Then with respect to coding right.

33
00:01:51,000 --> 00:01:53,000
Uh, cloudy 3.5 sonnet is the best.

34
00:01:53,000 --> 00:01:57,000
Then here you can also see GPT uh Gemini 1.5 Pro which is from Google.

35
00:01:57,000 --> 00:01:57,000
Right.

36
00:01:57,000 --> 00:01:59,000
So many companies.

37
00:01:59,000 --> 00:01:59,000
Right.

38
00:01:59,000 --> 00:02:00,000
See over here.

39
00:02:00,000 --> 00:02:02,000
And this is nothing but anthropic here.

40
00:02:02,000 --> 00:02:07,000
Also you can probably see uh, different different models are specifically there right over here.

41
00:02:07,000 --> 00:02:07,000
Right.

42
00:02:07,000 --> 00:02:13,000
And, uh, everybody's competing to probably get the best model out there here.

43
00:02:13,000 --> 00:02:18,000
Also, you can see with respect to quality and output speed, you'll be able to see Gemini 1.5 flash,

44
00:02:18,000 --> 00:02:19,000
right?

45
00:02:19,000 --> 00:02:23,000
Um, if the output is very high, but the quality is a little bit less right.

46
00:02:24,000 --> 00:02:28,000
Uh, if I probably consider GPT four or, you know, if, let's say the output speed is somewhere around

47
00:02:28,000 --> 00:02:33,000
90 and your quality general index, uh, is very good over here.

48
00:02:33,000 --> 00:02:35,000
The value is 100, right?

49
00:02:35,000 --> 00:02:39,000
With respect to GPT four turbo, the speed is less, but the quality is really good.

50
00:02:39,000 --> 00:02:42,000
So this is specifically with respect to the comparison that you see.

51
00:02:42,000 --> 00:02:47,000
Quality versus price here we already saw Claude three opus was really high.

52
00:02:47,000 --> 00:02:49,000
GPT four turbo is somewhere here and remaining.

53
00:02:49,000 --> 00:02:52,000
All the models are very, very less with respect to the price.

54
00:02:52,000 --> 00:02:57,000
But the quality three quality I find out this cloudy 3.5 sonnet is best.

55
00:02:57,000 --> 00:03:02,000
Then you have other models like Gemini 1.5 Pro and all right then output speed.

56
00:03:02,000 --> 00:03:04,000
You can see Gemini 1.5.

57
00:03:04,000 --> 00:03:10,000
Flash is basically having 165 tokens per second, uh, you know, over here.

58
00:03:10,000 --> 00:03:15,000
And then, uh, you can probably go ahead and compare with all the other models over here, right.

59
00:03:15,000 --> 00:03:17,000
Pricing input and output prices.

60
00:03:17,000 --> 00:03:20,000
Also, you can probably see, see whenever we have this construct kind of models.

61
00:03:20,000 --> 00:03:20,000
Right.

62
00:03:20,000 --> 00:03:23,000
These are basically fine tuned models that are available over here.

63
00:03:24,000 --> 00:03:26,000
So you can just go ahead and compare this.

64
00:03:26,000 --> 00:03:27,000
These are some additional information.

65
00:03:27,000 --> 00:03:31,000
And if you just want to see that, how many different types of supported models are there.

66
00:03:31,000 --> 00:03:36,000
Here you have GPT four oh by OpenAI turbo, GPT 43.5, turbo 3.5.

67
00:03:36,000 --> 00:03:41,000
Then with respect to Google, you have Google one point, Germany 1.5, Pro 1.5 flash gamma two.

68
00:03:41,000 --> 00:03:43,000
Gamma two is an open source model.

69
00:03:43,000 --> 00:03:47,000
See here you can see the difference between proprietary and open right?

70
00:03:47,000 --> 00:03:51,000
When we say proprietary, this is basically a paid model.

71
00:03:51,000 --> 00:03:54,000
When it is open it is an open source model over here right.

72
00:03:54,000 --> 00:03:57,000
Gamma seven being struct is a fine tuned of gamma model.

73
00:03:57,000 --> 00:03:59,000
So it is also open source.

74
00:03:59,000 --> 00:04:03,000
Then you have this meta meta basically comes up with all the model as open source.

75
00:04:03,000 --> 00:04:05,000
So here you can see all the open source model.

76
00:04:05,000 --> 00:04:06,000
Then you have Mistral.

77
00:04:06,000 --> 00:04:08,000
Uh two models are open and remaining.

78
00:04:08,000 --> 00:04:09,000
All are paid.

79
00:04:09,000 --> 00:04:11,000
Then you have this anthropic.

80
00:04:11,000 --> 00:04:15,000
All are paid cohere, two are open and two are paid.

81
00:04:15,000 --> 00:04:18,000
Then you also have this open chat.

82
00:04:18,000 --> 00:04:19,000
This is open right now.

83
00:04:19,000 --> 00:04:20,000
Databricks is also open.

84
00:04:20,000 --> 00:04:21,000
Rekha.

85
00:04:21,000 --> 00:04:24,000
Here you can see it is propriety and all the other things.

86
00:04:24,000 --> 00:04:26,000
Also, you can probably go ahead and check it.

87
00:04:26,000 --> 00:04:32,000
So a good handy website to probably go ahead and check the performance of each and every model.

88
00:04:32,000 --> 00:04:34,000
So I hope you like this particular video.

89
00:04:34,000 --> 00:04:35,000
Uh, I will see you all in the next video.

90
00:04:35,000 --> 00:04:36,000
Thank you.