OpenAI, the company that rose to fame with its state-of-the-art AI-powered chatbot ChatGPT has once again surprised the industry with its latest power-packed model of an AI-based image and text-understanding model known as Chat GPT-4. Although it can take inputs in both textual and visual form, the output will be purely text-based.
On a side note, the image input feature is yet to be launched publicly. For now, it’s only available to a single tester— Be My Eyes, an app developed to help visually impaired individuals with their day-to-day activities.
OpenAI also confirmed that this latest version has done away with most of the bugs and shortcomings in the previous version. It’s intuitive, creative, and collaborative. In addition to that, GPT-4 is far more capable of solving complex problems than all the other previous versions.
Despite such tremendous developments, a few minor problems of the system remain the same. For instance, GPT-4 is still prone to making up false information. The company warns that the bot isn’t aware of any event that happened after September 2021. Any questions after this timeline can confuse the program. However, this minor hiccup doesn’t seem to have hindered GPT-4’s demand.
Pricing of Chat GPT 4
The company has decided to set a usage limit for these users until the program is launched full-fledged. The pricing so far seems quite reasonable. The company has decided to charge $0.03 per 1,000 prompt tokens, which translates to roughly 750 words. Whereas the completion tokens come at a price of $0.06 per 1,000 tokens.
The prime difference between prompt tokens and completion tokens is that the prompt tokens are the words that you enter into GPT-4, whereas the completion tokens are the words of the final content that the software generates for you.
It’s also important to note that tokens are raw texts. For example, if you want to enter the word “interesting”, it will be broken into its syllables which are “in-te-res-ting”. So for one word, you’ll need four tokens.
How is Chat GPT 4 better?
OpenAI further warns that regular users might not be able to notice the difference between GPT-4 and its predecessors. However, the difference has been clearly recorded during the testing phase.
For instance, this latest version is way more prepared to sit for exams like SAT Evidence-Based Reading & Writing exams, Uniform Bar Exam, LSAT, SAT Math, etc. In all these exams, GPT-4 managed to get a score of 88 percentile or more.
OpenAI has already partnered with several companies, such as Khan Academy, Duolingo, and Stripe, over their latest launch.
While for most of us, GPT-4 might seem like a new launch; it has actually been in the market for quite some time now. As Microsoft confirms, its latest launched BingChat, which has been developed in collaboration with OpenAI, is actually running on GPT-4.
Share
Krishi Chowdhary
Journalist
