I think you are correct. A token according to open ai is 1 token ~= 4 chars in English 1 token ~= ¾ words 100 tokens ~= 75 words a character is only a single letter. Price difference token to charakter 0.01 / 0.00025 = 40 This means Gemini Api is roughly 10x cheaper then Open Ai. However if Gemini Pro is significantly worse compared to GPT4 you might want to use booth APIs and switch between them depending on how importend a good output is for you usecase.
@@VR_Wizard I afraid there is a mistake in your calculation. I havent found official OpenAI measurments for their tokens, but seen unofficial claims that for English text 1 ChatGPT token ≈ 3/4 words on average. An English word on average is 4.7 characters. Therefore 1 ChatGPT token ≈ 3.5 characters. The pricing difference per character: 0.001/3.5=0.000285 (GPT) vs 0.00025 (Gemini) Given the above measurements one could say that the pricing is similar if not slightly more pricier for ChatGPT. This is only for English texts, because other languages could have other tokenization rates and avg. word lengths.
@@alexxx4434 The price for a token (Input) is 0.01 not 0.001 for open AI. The 40 I got means a character in the google API costs 40x less then a token from OpenAI. Since OpenAI claims that their tokens contain 4 chatacters I end up with 10x price difference. The numbers are fom the help site of OpenAI "What are tokens and how to count them?" I cant post the link here my comment gets blocked otherwise.
Brother, Google is pay per character, which is worse than pay by tokens. So gpt is still cheaper than Gemini Because gpt is pay by 1k tokens not characters
Hi. Really an amazing content and it's working very well. But I got stuck somewhere, could you please try to make a video on how we can implement a multiple input images at a time and run our prompt on them For Eg: "Explain these 10 Images?". I couldn't able to figure out how it can achievable and it would be really grateful if you could ans this. Thanks in advance.
However that model equal other models do not math operation, with large operation always fails. For example: What is the sum total of: 2317+2446+2667+2270+2435+2419+2395+2438+2363+2427+2271+2219+2141+2326+2647+2446+2347+2535+2397
I have a question, if you please. I created 2 different Prompts in AI Studio: ChatbotBasketball and ChatbotSoccer, and have uploaded unique documents to each. I have created 2 API keys, how do I assign an API to a particular prompt, so that 1 key is just for the Basketball prompt and one for the Soccer prompt? I understand how to add the API code to an .env file, so that it works over the internet, but there is no step that sets the API to either of the Prompt Projects in AI studio. Please help!
Just to clarify, the API keys correctly connect to Google Cloud Projects, so the billing and traffic metrics work for each API key. However, neither API key / Prompt combo has access to the files uploaded to the Prompts in Google AI Studio. So the Prompt User / Model roles from Prompt questions are in the JavaScript, but the AI can't access the files / documents that are in the AI studio. The Prompt can only read the files if you are asking it questions directly in AI Studio..
i have tested gemini pro api but is not get results as bard direct i have checked some text generating by bard and other with gemini pro the gemini pro always dedicted by ai content
Great work! I am trying to figure out how to include the prompt data I generated on Google ai studio and exported to Google drive into the colab notebook tutorial. I am trying the multi turn conversation section in the notebook but no progress yet. Any thoughts? I am a novice btw.
Just using Google Bard with the image upload. I found that it couldn't do the following things when asked a basic question. What is the size of this image in pixels? Please give me the pixel location of the red button in the middle? Please tell me how many objects are in this image (simple white background with 3 geometric shapes of different primary colors). Is there some way to ask the question in a better way to get better answers from Gemini Vision Pro (which I'm assuming Google Bard is using)?
also youa re dumb to assume the model can work or analyze image in a "live" way like nofo pixel size background etc. the model takes in data from the image in a total or whole to analyze the image. no model can do what you are intending to do
Need help with Building your LLM powered Applications? Reach out if you need help: calendly.com/engineerprompt/consulting-call
"Impressive demo! Can't wait to try Gemini Pro API."
The free low count is a blessing for experimenting!
It's nice to have an option to choose.
I really love the content that you are providing..big thanks bro ❤
1:52 It appears that you are incorrectly comparing pricing for tokens (GPT) to characters (Gemini).
I think you are correct.
A token according to open ai is
1 token ~= 4 chars in English
1 token ~= ¾ words
100 tokens ~= 75 words
a character is only a single letter.
Price difference token to charakter
0.01 / 0.00025 = 40
This means Gemini Api is roughly 10x cheaper then Open Ai. However if Gemini Pro is significantly worse compared to GPT4 you might want to use booth APIs and switch between them depending on how importend a good output is for you usecase.
@@VR_Wizard I afraid there is a mistake in your calculation. I havent found official OpenAI measurments for their tokens, but seen unofficial claims that for English text 1 ChatGPT token ≈ 3/4 words on average. An English word on average is 4.7 characters. Therefore 1 ChatGPT token ≈ 3.5 characters.
The pricing difference per character: 0.001/3.5=0.000285 (GPT) vs 0.00025 (Gemini)
Given the above measurements one could say that the pricing is similar if not slightly more pricier for ChatGPT.
This is only for English texts, because other languages could have other tokenization rates and avg. word lengths.
@@alexxx4434 The price for a token (Input) is 0.01 not 0.001 for open AI. The 40 I got means a character in the google API costs 40x less then a token from OpenAI. Since OpenAI claims that their tokens contain 4 chatacters I end up with 10x price difference.
The numbers are fom the help site of OpenAI "What are tokens and how to count them?" I cant post the link here my comment gets blocked otherwise.
OpenAi charges per 1k tokens where as Gemini charges per 1k char. Could you please compare the price based on this?
a token is 4 chars of base64 that equils 3 letters
The prompt_feedback property gives the result of the prompt or the generated response.
I really thank you for your video
Glad it was helpful
Brother, Google is pay per character, which is worse than pay by tokens. So gpt is still cheaper than Gemini Because gpt is pay by 1k tokens not characters
I did a calculation in my other comment, converting tokens to chars. It appears that the pricing could be actually on par.
Hey, thank you very much !
Hi. Really an amazing content and it's working very well. But I got stuck somewhere, could you please try to make a video on how we can implement a multiple input images at a time and run our prompt on them For Eg: "Explain these 10 Images?". I couldn't able to figure out how it can achievable and it would be really grateful if you could ans this. Thanks in advance.
Really useful video, thank you. And it's great to have some free tokens to test the model.
However that model equal other models do not math operation, with large operation always fails. For example: What is the sum total of: 2317+2446+2667+2270+2435+2419+2395+2438+2363+2427+2271+2219+2141+2326+2647+2446+2347+2535+2397
Sir I want a chat application but i want to fine-tune my model for that task how can i do IT. Example a bot for agriculture
You can use Openai Assistant API
Can you explain the process, how to get early access of gemini and how you were able to get it,
I have a question, if you please. I created 2 different Prompts in AI Studio: ChatbotBasketball and ChatbotSoccer, and have uploaded unique documents to each. I have created 2 API keys, how do I assign an API to a particular prompt, so that 1 key is just for the Basketball prompt and one for the Soccer prompt? I understand how to add the API code to an .env file, so that it works over the internet, but there is no step that sets the API to either of the Prompt Projects in AI studio. Please help!
Just to clarify, the API keys correctly connect to Google Cloud Projects, so the billing and traffic metrics work for each API key. However, neither API key / Prompt combo has access to the files uploaded to the Prompts in Google AI Studio. So the Prompt User / Model roles from Prompt questions are in the JavaScript, but the AI can't access the files / documents that are in the AI studio. The Prompt can only read the files if you are asking it questions directly in AI Studio..
i have tested gemini pro api but is not get results as bard direct i have checked some text generating by bard and other with gemini pro the gemini pro always dedicted by ai content
When will paid version be available?
Unfortunately, I'm getting "Internal Server Error" on Google Colab... ERROR:tornado.access:500 a GET error
How can we fine tune for a task
bro make a video on how to use gemini api in metagpt !!
Thanks but I just wanted the video to start with the most simple option of CHAT accessing the API FROM ANY PROGRAM using REST ( HTTP GET). Any ideas ?
Also to act like a chatbot do I need to prefix every question with "In one sentence" ? Any ideas ?
Check out my latest video. Hope that will be useful.
It does not work in EU.
Great work! I am trying to figure out how to include the prompt data I generated on Google ai studio and exported to Google drive into the colab notebook tutorial. I am trying the multi turn conversation section in the notebook but no progress yet. Any thoughts? I am a novice btw.
"Great tutorial, I'm excited to try Gemini Pro API on AI Studio!"
Just using Google Bard with the image upload. I found that it couldn't do the following things when asked a basic question. What is the size of this image in pixels? Please give me the pixel location of the red button in the middle? Please tell me how many objects are in this image (simple white background with 3 geometric shapes of different primary colors). Is there some way to ask the question in a better way to get better answers from Gemini Vision Pro (which I'm assuming Google Bard is using)?
google bard doesnt use the gemini vision.
also youa re dumb to assume the model can work or analyze image in a "live" way like nofo pixel size background etc. the model takes in data from the image in a total or whole to analyze the image. no model can do what you are intending to do
is the model available also in spanish?
I believe so too. Europe might not have access yet though
Cannot make make API
Can I use Gemini API key in my vscode??
it depends on your location, right now you can't from Europe
Yes if you have access to it.
Well .. learn Chinese guys.
Why would anyone care? They basically lied to make themselves seem relevant and GPT 3.5 and Mixtral are better.
Access restricted
You do not have permission to view this page.
Are you having trouble accessing the colab or Google AI studio?
@@engineerpromptGoogle ai studio
Hey, thank you very much !