GPT Tokens to Cost Calculator
Total Cost:
$0.00
Select a Model and Enter Tokens
Embedding Models Tokens to Cost Calculator
Total Cost:
$0.00
Select a Model and Enter Tokens
Fine-tuning Models Cost Calculator
Total Cost:
$0.00
Select a Model and Enter Tokens
Audio Models Cost Calculator
Total Cost:
$0.00
Select a Model and Enter Input
Frequently Asked Questions
1. How are the costs for language model usage calculated?
The costs are calculated based on the number of input and output tokens processed
by the model. The price is set per million tokens, so the formula used is:
total cost = (input token cost * number of input tokens + output token cost * number of output tokens)
/ 1,000,000. For example, GPT-4o charges $2.50 per million input tokens and $10.00 per million output tokens.
2. What's the difference between the latest GPT models?
The latest models include GPT-4.5 Preview which costs $75/1M input and $150/1M output tokens, GPT-4o which costs $2.50/1M input and $10.00/1M output tokens, and GPT-4o Mini which offers a more cost-effective option at $0.15/1M input and $0.60/1M output tokens. The o1 and o1-pro models offer higher capabilities at $15/1M input, $60/1M output and $150/1M input, $600/1M output respectively.
3. How have the embedding model options changed?
There are three main embedding models: text-embedding-3-small ($0.02/1M tokens), text-embedding-3-large ($0.13/1M tokens), and text-embedding-ada-002 ($0.10/1M tokens). The -3 models are newer and offer improved performance, with the small version being more cost-effective and the large version offering higher quality embeddings.
4. What are the costs involved in fine-tuning models?
Fine-tuning costs vary significantly between models. GPT-4o fine-tuning costs $25/1M tokens for training, $3.75/1M for input, and $15/1M for output. GPT-3.5 Turbo is more economical at $8/1M for training, $3/1M for input, and $6/1M for output. The GPT-4o Mini option provides a middle ground at $3/1M for training.
5. What audio and speech models are available?
Audio models include GPT-4o Mini TTS ($0.015/minute), GPT-4o Transcribe ($0.006/minute), and GPT-4o Mini Transcribe ($0.003/minute). Text-to-speech models include TTS ($15/million characters) and TTS HD ($30/million characters) for higher quality speech generation.
6. How do the new GPT-3.5 Turbo prices compare to older versions?
The latest GPT-3.5 Turbo (0125) has been optimized for better cost efficiency at $0.50/1M input and $1.50/1M output tokens. This is more economical than previous versions which charged $1.00-$1.50/1M for input and $2.00/1M for output tokens. The 16k context version costs $3.00/1M input and $4.00/1M output tokens.
7. What are o1 and o1-pro models?
The o1 ($15/1M input, $60/1M output) and o1-pro ($150/1M input, $600/1M output) are newer high-performance models designed for advanced tasks requiring greater capabilities. They represent the premium tier of language models with significantly higher pricing reflecting their enhanced capabilities.
8. How are audio model costs calculated?
Audio model costs are typically calculated per minute of audio for transcription services (like Whisper at $0.006/minute) or per million characters for text-to-speech services (like TTS at $15/million characters). The GPT-4o audio models combine language understanding with audio processing capabilities.
9. How can I optimize costs with the new pricing structure?
Consider using GPT-4o Mini for tasks that don't require full GPT-4o capabilities, or GPT-3.5 Turbo for simpler tasks. For embeddings, text-embedding-3-small offers good performance at a lower cost. When fine-tuning, carefully evaluate whether you need GPT-4o's capabilities or if GPT-3.5 Turbo would suffice, as the cost difference is significant.
10. What's the difference between GPT-4o and GPT-4 Turbo?
GPT-4o ($2.50/1M input, $10/1M output) represents a newer generation of models with improved capabilities at a significantly lower price point than GPT-4 Turbo ($10/1M input, $30/1M output). GPT-4o offers better performance while being approximately 4 times less expensive, making it the recommended choice for most applications.