June 15, 2023
Hello there, I am Hiroshi Nishio, the spokesperson for 'Q, ChatGPT for Slack'. We have some thrilling news to share! OpenAI has recently introduced the GPT-3.5 16K model on June 13, 2023, and we're excited to announce that we'll be adding a new plan supporting this model to 'Q, ChatGPT for Slack' very soon.
GPT-3.5 16K handles 4 times the token limit of the previous GPT-3.5 (4K) model, and although it comes with double the API usage cost, the benefits are definitely worth it. 😃
If you're unsure what a token is, think of it as a piece of text. The length of a token can vary depending on the language, but in English, a token typically corresponds to around 4 characters or 1 word. For Japanese, a token often equates to fewer than 1 character. Here's an official explanation from OpenAI about tokens.
Tokens serve as a limit for the combined total of questions and answers. Thus, the higher the token limit, the longer the questions and answers you can get.
To put this into perspective, until now, if we wanted to reserve around 2,700 tokens for the conversation history including the question and about 800 for the output, we could only handle about 500 tokens for hints. Considering that an answer is usually found within the top 5 results of a Google search, this meant that we could only handle around 100 tokens (or 400 characters in English and less than 100 characters in Japanese) per hint.
But this is about to change.
For the previous GPT-3.5 model, the token limit was 4,000, allowing for roughly 16,000 characters in English. However, with the new GPT-3.5 16K, we're looking at a significant boost up to 64,000 characters! 😮
With 64,000 characters at your disposal, 'Q, ChatGPT for Slack' is set to deliver enhanced memory and generate longer outputs in one go. Whether it's producing multilingual blog post drafts or dealing with entire academic papers, GPT-3.5 16K has got it covered.
The news is even better for our plugins. 'Q, ChatGPT for Slack' supports plugins just like the original ChatGPT, with popular examples including WEB search and PDF search.
When 'Q' determines it cannot answer a user's query, it leverages these plugins to seek relevant reference information, just like a Google search, to assist in formulating a response. And since the token limit includes these hints, the new model's increased token limit means we can handle more detailed hints. 📊
Words can hardly express how thrilled we are to see the incredible potential of the GPT-3.5 16K model. We're excitedly preparing to introduce the new plan, so stay tuned for updates.
In the meantime, don't forget to explore the power of 'Q, ChatGPT for Slack' through our homepage, blog, or trial installation. And if you're ready to take the plunge, check out our pricing and subscription details.
We're eager to welcome you to the new horizon of AI intelligence with 'Q, ChatGPT for Slack'. Here's to a more intelligent, efficient, and seamless future! 🥂
14-days free trial.
No credit card required.
The first day of the trial is GPT-4!