#Calculations with prices from openAI

1 messages · Page 1 of 1 (latest)

thorny zephyr
#

say I send about 100 messages per hour with an average token count of 125 per message? How much does this cost me per month with the different models for chat completion?

grave crestBOT
#

Hi @thorny zephyr. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

To calculate the cost, you need to know the number of tokens used and the cost per token for the specific model you're using.

Let's break it down:

  1. You send 100 messages per hour, each with an average of 125 tokens. So, you're using 12,500 tokens per hour.
  2. In a day (24 hours), you'll use 12,500 tokens/hour * 24 hours/day = 300,000 tokens/day.
  3. In a month (assuming 30 days), you'll use 300,000 tokens/day * 30 days/month = 9,000,000 tokens/month.

Now, let's calculate the cost for different models:

  1. For gpt-3.5-turbo-1106, the cost is $0.0010 per 1K tokens for input and $0.0020 per 1K tokens for output. Assuming a 50-50 split between input and output tokens, the average cost per token is ($0.0010 + $0.0020) / 2 = $0.0015 per token. So, the monthly cost would be 9,000,000 tokens/month * $0.0015/token = $13,500/month.

  2. For gpt-3.5-turbo, the cost is $0.0015 per 1K tokens for input and $0.0020 per 1K tokens for output. Assuming a 50-50 split between input and output tokens, the average cost per token is ($0.0015 + $0.0020) / 2 = $0.00175 per token. So, the monthly cost would be 9,000,000 tokens/month * $0.00175/token = $15,750/month.

  3. For gpt-4 with 128k context lengths, the cost is $0.01 per 1K prompt tokens and $0.03 per 1K sampled tokens. Assuming a 50-50 split between prompt and sampled tokens, the average cost per token is ($0.01 + $0.03) / 2 = $0.02 per token. So, the monthly cost would be 9,000,000 tokens/month * $0.02/token = $180,000/month.

Please note that these calculations are approximations and the actual cost may vary depending on the exact distribution of input and output tokens and the specific models used. You can check the exact token usage and cost in the usage field in the API response and on the OpenAI Pricing page (https://openai.com/pricing).

#

@thorny zephyr kapa.ai is still learning and improving, please let us know how it did by reacting below

thorny zephyr
#

@grave crest alright, let's say I have a budget of 50 dollar how many messages can I send per hour?

grave crestBOT
#

Let me check the knowledge base for you. One moment! :books:

#

To calculate the number of messages you can send within your budget, you need to know the cost per token for the specific model you're using and the average number of tokens per message.

Let's break it down:

  1. Let's assume you're using the gpt-3.5-turbo model, which costs $0.0015 per token on average (assuming a 50-50 split between input and output tokens as explained in the previous response).

  2. If each message contains an average of 125 tokens, then the cost per message is 125 tokens/message * $0.0015/token = $0.1875/message.

  3. With a budget of $50, you can send $50 / $0.1875/message = approximately 267 messages.

  4. If you want to distribute these messages evenly over an hour, you can send 267 messages/hour.

Please note that these calculations are approximations and the actual number of messages may vary depending on the exact distribution of input and output tokens and the specific models used. You can check the exact token usage and cost in the usage field in the API response and on the OpenAI Pricing page (https://openai.com/pricing).

Sources:

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@thorny zephyr kapa.ai is still learning and improving, please let us know how it did by reacting below

thorny zephyr
#

@grave crest so how many can I send per hour on a monthly basis? with 50 dollar

grave crestBOT
#

Let me check the knowledge base for you. One moment! :books:

#

To calculate the number of messages you can send per hour on a monthly basis within your budget, you need to know the cost per token for the specific model you're using and the average number of tokens per message.

Let's break it down:

  1. Let's assume you're using the gpt-3.5-turbo model, which costs $0.0015 per token on average (assuming a 50-50 split between input and output tokens as explained in the previous response).

  2. If each message contains an average of 125 tokens, then the cost per message is 125 tokens/message * $0.0015/token = $0.1875/message.

  3. With a budget of $50, you can send $50 / $0.1875/message = approximately 267 messages.

  4. If you want to distribute these messages evenly over a month (assuming 30 days), you can send 267 messages / 30 days = approximately 9 messages per day.

  5. If you want to distribute these messages evenly over an hour (assuming 24 hours in a day), you can send 9 messages/day / 24 hours/day = approximately 0.375 messages per hour.

This means, with a budget of $50 per month, you can send approximately 1 message every 2-3 hours using the gpt-3.5-turbo model.

Please note that these calculations are approximations and the actual number of messages may vary depending on the exact distribution of input and output tokens and the specific models used. You can check the exact token usage and cost in the usage field in the API response and on the OpenAI Pricing page (https://openai.com/pricing).

#

@thorny zephyr kapa.ai is still learning and improving, please let us know how it did by reacting below

thorny zephyr
#

@grave crest Ok, let's say we don't send messages during the night. So from midnight to about 7 in the morning. How many can I send per hour in a month

grave crestBOT
#

Let me check the knowledge base for you. One moment! :books:

#

To calculate the number of messages you can send per hour on a monthly basis within your budget, excluding the night hours, you need to know the cost per token for the specific model you're using and the average number of tokens per message.

Let's break it down:

  1. Let's assume you're using the gpt-3.5-turbo model, which costs $0.0015 per token on average (assuming a 50-50 split between input and output tokens as explained in the previous response).

  2. If each message contains an average of 125 tokens, then the cost per message is 125 tokens/message * $0.0015/token = $0.1875/message.

  3. With a budget of $50, you can send $50 / $0.1875/message = approximately 267 messages.

  4. If you want to distribute these messages evenly over a month (assuming 30 days), you can send 267 messages / 30 days = approximately 9 messages per day.

  5. If you want to distribute these messages evenly over the day, excluding the night hours (from midnight to 7 AM, so you have 17 hours of operation), you can send 9 messages/day / 17 hours/day = approximately 0.53 messages per hour.

This means, with a budget of $50 per month, you can send approximately 1 message every 2 hours during the operational hours (7 AM to midnight) using the gpt-3.5-turbo model.

Please note that these calculations are approximations and the actual number of messages may vary depending on the exact distribution of input and output tokens and the specific models used. You can check the exact token usage and cost in the usage field in the API response and on the OpenAI Pricing page (https://openai.com/pricing).

#

@thorny zephyr kapa.ai is still learning and improving, please let us know how it did by reacting below

thorny zephyr
#

@grave crest Alright, let's decrease our operational hours. We are only doing 5 PM to 11 PM

grave crestBOT
#

Let me check the knowledge base for you. One moment! :books:

#

To calculate the number of messages you can send per hour on a monthly basis within your budget, during the specific operational hours, you need to know the cost per token for the specific model you're using and the average number of tokens per message.

Let's break it down:

  1. Let's assume you're using the gpt-3.5-turbo model, which costs $0.0015 per token on average (assuming a 50-50 split between input and output tokens as explained in the previous response).

  2. If each message contains an average of 125 tokens, then the cost per message is 125 tokens/message * $0.0015/token = $0.1875/message.

  3. With a budget of $50, you can send $50 / $0.1875/message = approximately 267 messages.

  4. If you want to distribute these messages evenly over a month (assuming 30 days), you can send 267 messages / 30 days = approximately 9 messages per day.

  5. If you want to distribute these messages evenly over the day, during the specific operational hours (from 5 PM to 11 PM, so you have 6 hours of operation), you can send 9 messages/day / 6 hours/day = approximately 1.5 messages per hour.

This means, with a budget of $50 per month, you can send approximately 1-2 messages every hour during the operational hours (5 PM to 11 PM) using the gpt-3.5-turbo model.

Please note that these calculations are approximations and the actual number of messages may vary depending on the exact distribution of input and output tokens and the specific models used. You can check the exact token usage and cost in the usage field in the API response and on the OpenAI Pricing page (https://openai.com/pricing).

#

@thorny zephyr kapa.ai is still learning and improving, please let us know how it did by reacting below

thorny zephyr
#

@grave crest Alright, lets change it up 7 AM to 11 PM only on friday, saturday and sunday.

grave crestBOT
#

Let me check the knowledge base for you. One moment! :books: