OpenAI's Official ChatGPT API and Whisper API - Pricing & Access Tutorial

Key Takeaways

hey friends welcome to one little coder one of the most avoided news of this year probably is chargpt APA and today chargpt API and Whisper API are both available on open AI development platform yes if you are a developer and you wanted to use charge apt or you wanted to use whisper API as an API within your application starting today you can do that and the surprising thing is the cost and that we are going to discuss in this video the way they have priced this is quite amazing I'm not sure how open a is going to make money out of this but for now let's enjoy while it is amazing so in this video I'm going to show you what is the announcement that opening is made in terms of charge GPT API and Whisper API then we are going to look at how to access charge GPD API and also how to access whisper API and then finally we are going to discuss the pricing aspect of how chart GPT API and Whisper API are placed and at the end there are certain developer friendly changes is that open EA has made with respect to their apis we are going to discuss that and wrap the video so let's get started the first thing is as you can see introducing charge epd and Whisper apis developers can now integrate chat GPT and Whisper models into their applications and products through our API how do you access the API go to platform.openai.com platform.openai.com you can access it but I'll show you later if you see that open a is claiming that they have managed to reduce the cost of charge GPT by 90 percent since December so they're going to use that savings and fuel as API so they're going to give that savings to the API users we don't know if this 90 is coming from people who are paying for chargpt or we don't know if it is just simply like open AI has done optimizations to reduce it but yeah they've said optimization but that's that's what they have saying so now also on the whisper aspect the large E2 model which is the latest large model the that has got a lot of improvements I have had a different video about the large V2 model LED so that model whisper model which is a speech to text model so speech to text model and charge gpt's chatbot model or both available as API within open AI developer platform now they have given some of examples like how people are using it so what kind of things people have been doing with chargpt apis and Whisper APA one is an example with Snapchat another is an example I mean like in fact like with the Snapchat news everybody was shocked thinking that Snapchat actually launched its own chatbot but today you get to know that it's not charged up it's not Snapchats or snaps own chatbot but indeed they are leveraging charge gbt API and also you can see this instacart demo so where instacart is using charge gbt API to you to show the products that are relevant to Instagram link if Instagram has got certain products so you can ask question using charge GPT and they are actually using charge apt API and you can ask version and it's going to give you responses using charge GPT APA but within their products so I don't know how they are doing it but other than fine tuning a model but this is what it is that's amazing now you have got a bunch of other more examples for you to see different languages and all these examples now how do you access chatgpt APA a lot of people know that gpt3 was a model GPT 3.5 was a model but other than developer Community people wouldn't know the model name in itself so if you know gpt3 you would have known that it is text Darwin CO2 and if you had known chat GPT 3.5 you would know that it's text Darwin c003 in the same way now the chart GPT model is called GPT 3.5 Turbo so what is this this is the same model that is used in the chat gbt product so if you go to chart.openea.com the model that is used in chat.openhair.com is the same as GPT 3.5 turbo that's the model now how is it priced it is quite surprising because the price is point zero zero to per thousand tokens so for every thousand token you're going to pay point zero zero two which which means like what you need at least 500 yeah 500 or 5000 so somewhere around like my math is not really good but uh you can see for you to even pay one dollar there are a lot of tokens that you can exchange now if you are not familiar with the token World token is approximately like one word so if you have thousand words you can say 750 tokens approximately so this is this is like an approximate calculation and this is quite amazing this is quite aggressively priced let me give you a context so if you go to the charge GPT or open AI pricing page you can see that the GPT 3.5 turbo model the charge GPT model which like literally everybody has been behind chargpt API is point zero zero two per thousand token now you might assume that you know what uh the GPT 3 should be cheaper than this because stat GPT is so famous no it's not in fact you would see that the DaVinci model which is one of their most powerful models The DaVinci family model is 0.002 which is 10 times more expensive than chart GPT or in other words I can say chat gpt's API cost is 10 times cheaper than down C now I'm not quite sure what context people are going to prefer DaVinci but as you can see like it's uh it is amazing it is like the way it has been priced is quite unbelievable why it is a massive deal you need to know it is a massive deal because people have been building GPT 3 based applications uh even um before text DaVinci over 3 was released but since the arrival of text DaVinci over 3 there has been a significant Improvement in the performance the cost was cut down previously it was uh it was a 0.006 Darwin C model cost was 0.006 dollars for 1000 tokens but open AI reduce the price from that to 0.002 and uh that that that made a lot of developers build applications on top of GPT 3.5 model that text 003 model and now you have got a much better model like everybody in the world is a fan of this model and that is going to be one tenth of this cost it's going to be 10 almost like 10 times cheaper than what this model which was already more powerful which was already like widely respected a lot of people were building applications on top of it so this is this is completely ridiculously aggressively priced I'm not sure what is the intention behind it but if I were to guess it maybe open air doesn't want uh Google to make any money maybe open AI doesn't want like anthropic the company that is building Claude maybe they don't want any of the competitors to be in the scene so they want to just know or they they already know that charge if it is quite popular so they just want to ride on the popularity and then make everybody use the API I'm not I'm not I'm not honestly sure what would people do after somebody increases in apa cost let's say you are one or two years down the line so it's completely possible that for one or two years um open AI is looking at this VC funded cash burning rate and then try to make money uh anyways this company was called open EA that's another topic altogether but it looks like open a can still make money after some time where people get used to it but at least at this point this is this is like the best time to build AI applications AI artificial intelligence applications because your chat GPT model GPT 3.5 turbo is it's ridiculously cheap and aggressively priced and this is this is an aggressive price so now that's that's uh about open AI is charging PT now how can you access charge GPT model so the way you can access charge GPT model is very simple so you can make a request like curl requires you can use request package in Python whatever you want to do you can make a request this is the APA endpoint open a.com slash V1 chat slash chat slash completions you need to have your opening API key and then send the question like whatever the question is and then you can send the message you can make a call to this model and what would be the response the response would look something like this here the role is the user and here the role is the assistant like if you are building a conversational agent one is the user the other one is the assistant so this way you can send conversational UI um conversational message and you can also see what kind of message it has got with all the details chat ID messages the model name what all these things so this is quite quite amazing and um and also like charge GPT itself has got an upgrade so what what they are saying is developers who use GPT 3.5 turbo model will always get the recommended stable model and you can also opt for specific models so for example today they have released a new version of the turbo model which is turbo zero three zero one and uh you know you can you can start accessing it so it looks like open AI wants to experiment a lot and by paying for the API GPD API you are going to get that privilege of getting uh getting exposure to what getting exposed to what they're experimenting and they have got some information about uh if you want to do it for Enterprise how you can do it through Azure and all this kind of information now that is chargpt API now let's talk about whisper like I still think that whisper is really one of the greatest thing that has come out of open AI a lot of people have been talking about charge GPT takes DaVinci understand but this is an open source Model A lot I have huge respect for this model given what it can do and today starting today whisper API is available inside open AI development platform now if you do not know anything about whisper I've got a lot of videos about whisper how to do normal translation normal translation how to do just transcription how to do uh speaker diarization I have all these videos please check it out but if you simply want to look at it now there is a new model the endpoint audio transcriptions and this model once again you need to have open a APA key and uh the only bummer here is that you need to send the entire audios payload so I don't know at least right now it doesn't look like they're supporting streaming audio which is quite a bummer because a lot of times people want to send streaming audio but in this case you need to send the entire audio clip in the payload and then you are going to get the response most likely they will definitely change or they might introduce another model with a higher pricing tier but at least at this point you can get transcriptions and you can get translations translations get translated into English transcription happens in The Source language like if I'm speaking in Tamil I'm going to get Tumblr but if I want translation then my thumbnail will get first translated and transcribed into Tamil text from there it gets translated into English so it accepts a wide variety of formats like most popular audio formats now looking at whisper pricing let's look at whisper pricing what does it say if you look at whisper pricing whisper says for one minute the pricing is 0.006 so what do you if you want point zero 6 into 60 minutes so if you're going to do it for 60 minutes like for one hour you're going to pay approximately let's say three cents that's quite good I don't I've not seen a lot of people uh already having whisper and protection at a good rate so for one for one hour audio clip you're going to pay three cent so this is uh this this looks 30 Cent I think this is 30 Cent yeah I'm sorry 30 Cent so this is good this is actually good and uh let's see what kind of things that people can do with this so now this is all about the cost I'm going to show you now how you can access the model all you have to do is you have to go to platform.openei.com once you go to platform.openea.com you can go to playground and once you go to playground you can select the desired model right now at least while I'm recording this video I don't see any chat GPT model in itself here but you can see every other model that is available and you have the small icon here that can help you record a audio and then you can upload so and again you have you have all this information and you can also click view the code and you can get the code this this can help you to get the code in Python JavaScript whatever language that you want this is one way of accessing the model to try the next important thing is you can go check the documentation when you go check the documentation click documentation click models here you will get all the model information so one you can see GPT 3.5 and inside that you can see what is GPT 3.5 turbo can do what is GPT 3.5 turbo 30301 can do and what is the maximum request that you can do what at till what level the training data is available September 2021 all these informations are available here the next thing is look at whisper once again you can see what is the model what is the model that is using uh whisper V2 large model and then you can see currently there is no difference between the open source version and version available through API so which means if you manage to use whisper CPP or something in your local machine that's good enough you don't need APA but yeah if you need a API you can see so this is this is all how you can access the API all you have to do is get the opening a API key and then you have got the curl response that you have to make direct sorry the code request that you have to make for both whisper and the chat jpt model which in this case is a GPT 3.5 turbo so now this covers everything that we wanted to discuss starting from what is available pricing how to access everything so just to quickly say how to access go to platform.openai.com yeah if you have never logged in most likely you would have then you need to sign up first after that you can go check everything that you want inside that platform now finally there is some interesting changes that opening has decided to do which they are calling developer Focus so over the past six months we have been collecting feedback from our APA Partners sorry APA customers to understand how we can serve them better and they've made concrete changes that's what they're saying data submitted through APA is no longer used for service improvements which means I think in playground or charge GPT if you write something that's always being used for model Improvement but if you're going to submit something through API that is not going to be used for service improvement no unless you opt in so by default you are opted out unless you opt in it's not even used for model training that's quite surprising implementing they're implementing a default so TDA retention policy for API users with options for stricter retention depending upon user need that's good and a lot of organizations have data retention policy so looks like they've got some feedback they're removing or removing our pre-launch review unlocking by or improving automated monitoring okay developer documentation Improvement okay then simplifying the terms of usage and okay I think the most important thing is how they are handling the data and then the data retention I think this might encourage some more companies to try out these apis especially with chargpt now available as an EPA in the form of GPD 3.5 turbo model so overall it looks like um open EA is a full-on uh want to be developer friendly cost efficient aggressive uh a very aggressive pricing and also it seems that they have really good plans how to grow this uh Fame around charge GPT or the buzzer on charge GPT and looks like this API couldn't have come out a bit of time so everybody has been looking for an official charging PT APA from open air and today you have the I I don't know I don't want you to forget that whisper AP is also available because everybody is going to talk about chat GPT but you have got a charge GPD API whisper API official charge GPT and Whisper APS from Hooper a I hope this was helpful to you in learning about open ai's new API enrollment if you have any question let me know in the comment section otherwise peace