The ChatGPT API offers significantly reduced pricing for developers seeking powerful language models. Compared to the text-davinci-003 model, the API is 10 times less expensive (90% off), making it more accessible to developers integrating natural language processing into their applications.
ChatGPT Plus API pricing
ChatGPT Plus is available as a monthly subscription for $20, excluding taxes. It is unclear whether OpenAI will maintain this price for ChatGPT Plus in the future.
Note that unlike using the API, ChatGPT Plus currently has a limit of 25 messages every 3 hours if you use the GPT-4 model.
ChatGPT Plus vs. ChatGPT API pricing comparison
With a budget of $20USD, you can process about 7.5 million words through the ChatGPT API, which is the equivalent of ~10 King James Bibles. To estimate the number of tokens in a text, you can use the OpenAI Tokenizer.
Why is GPT-4 not free?
GPT-4 requires high costs for hardware upgrades, updates, additional features, hosting, ChatGPT would cost OpenAI about $700,000 per day. It is therefore essential for the company to charge for the GPT-4 language model to cover expenses.
GPT-4 features and capabilities
- Reliable and accurate: GPT-4 is more reliable and accurate than other OpenAI models.
- Human-level performance: GPT-4 is capable of human-level performance on several academic and professional benchmarks.
- Handling task complexity: GPT-4 can handle complex and difficult tasks by reaching a sufficient threshold.
- Visual inputs: GPT-4 can accept image and text prompts, unlike previous models.
GPT-4 API Pricing Analysis
If you access GPT-4 via ChatGPT Plus, you will need to subscribe to the $20 monthly fee. If you access GPT-4 via the API, there are two pricing options available based on context lengths. OpenAI has reduced the price of prompt tokens to make it easier to use GPT-4.
Costs for models with 8K context lengths (gpt-4 and gpt-4-0314):
- Prompt tokens cost $0.03/1k
- Completion tokens $0.06/1k
Costs for models with context lengths of 32K (gpt-4-32k and gpt-4-32k-0314):
- Prompt tokens cost $0.06/1k
- Completion tokens cost $0.12/1k
*1,000 tokens correspond to about 750 words.
Compared to GPT-3.5, the GPT-4 API has doubled the maximum context length of tokens from 4096 to 8192. The default throughput limits for the GPT-4 API are 40,000 tokens per minute and 200 requests per minute.
To join the GPT-4 API waiting list, click here.
Cost comparison between the GPT-4 API and the ChatGPT API
With a budget of $20, you can process about 444,000 tokens (or 333,000 words) with the GPT-4 API, assuming an equal split of prompt and completion tokens. However, the most relevant comparison is between the GPT-4 API and the GPT-3.5 API:
- GPT-4 for prompts is 14 times more expensive than the GPT-3.5 API
- GPT-4 for completions is 29 times more expensive than the GPT-3.5 API
The question to ask is whether the GPT-4 API adds 14 to 29 times the value of the GPT-3.5 API for your use case. If you work in the legal field or education/tutoring in general, it probably makes sense to test GPT-4. For other use cases, extensive testing will be required to determine if the additional costs are justified over the ChatGPT API.
Conclusion
The choice between the ChatGPT API and the GPT-4 API will depend on a careful analysis of your project needs and constraints. When evaluating these leading technologies, consider the key factors:
- the intended application
- desired accuracy
- ethical considerations
- cost implications
- adaptability to future developments
Your final decision will reflect your vision and commitment to harnessing the revolutionary potential of artificial intelligence.
