OpenAI finally launched GPT-4, their much anticipated new image and text understanding AI model, which they describe as “the latest milestone in its effort in scaling up deep learning.”
OpenAI has made GPT-4 accessible through its API today, with a waitlist, and through ChatGPT Plus, a paid upgrade for its popular AI-powered chatbot.
here is GPT-4, our most capable and aligned model yet. it is available today in our API (with a waitlist) and in ChatGPT+.https://t.co/2ZFC36xqAJ
it is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it.
— Sam Altman (@sama) March 14, 2023
What is OpenAI GPT-4?
GPT-4 is OpenAI’s latest LLM. OpenAI claims that GPT-4 performs at the “human level” on various professional and academic criteria and can accept both image and text inputs, an upgrade over its predecessor, GPT-3.5, which only accepted text.
Announcing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment: https://t.co/TwLFssyALF pic.twitter.com/lYWwPjZbSg
— OpenAI (@OpenAI) March 14, 2023
It took OpenAI six months of iterative alignment using lessons from an adversarial testing program and ChatGPT to get “best-ever results” on factuality, steerability, and refusing to step beyond of guardrails, the company claims.
“In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle,” OpenAI wrote in a blog post announcing GPT-4. “The difference comes out when the complexity of the task reaches a sufficient threshold — GPT-4 is more reliable, creative and able to handle much more nuanced instructions than GPT-3.5.”
According to OpenAI’s introduction of GPT-4, the system is “82 percent less likely to respond to requests for disallowed content and 40 percent more likely to produce factual responses than GPT-3.5.” This improvement is the result of six months of safety training.
GPT-4 vs GPT-3: Biggest differences
Even though GPT-4 is multimodal, it does so in fewer channels than was initially anticipated. According to OpenAI, the system can take in text and images and produce written results.
OpenAI claims it can understand more nuanced data since it can analyze text and images simultaneously. You may see how the algorithm describes memes and peculiar images in the examples provided below:
Analyzing text and images simultaneously is not the only difference. When compared to ChatGPT, GPT-4’s superior reasoning abilities put it in the lead.
What GPT can do is not finished yet. In summary, these are what sets GPT-4 apart:
- More precise
- Improved accuracy
- Describing images with impressive detail
- It has a sense of humor
- It can solve puzzles.
- It can ace standardized tests.
However, spoiler alert, it is still hallucinating.
How to use GPT-4?
To use GPT-4, you must meet at least one of the following requirements:
- API Waitlist: Sign up for GPT-4 waitlist to get rate-limited access to the GPT-4 API – which uses the same ChatCompletions API as gpt-3.5-turbo. OpenAI will start inviting some developers today and scale up availability and rate limits gradually to balance capacity with demand.
- Priority Access: Developers can get prioritized API access to GPT-4 for contributing model evaluations to OpenAI Evals that get merged.
- ChatGPT Plus: ChatGPT Plus subscribers will get GPT-4 access on chat.openai.com with a dynamically adjusted usage cap. The usage cap will depend on demand and system performance. API access will still be through the waitlist.
GPT-4 API pricing
- GPT-4 with an 8K context window (about 13 pages of text) will cost $0.03 per 1K prompt tokens, and $0.06 per 1K completion tokens.
- GPT-4-32k with a 32K context window (about 52 pages of text) will cost $0.06 per 1K prompt tokens, and $0.12 per 1K completion tokens.
- Overview page of GPT-4 and what early customers have built on top of the model.
- Blog post with details on the model’s capabilities and limitations, including eval results.
Is GPT-4 good? Let’s test it
We weren’t kidding when we said let’s test. Here are the test results of the GPT-4:
GPT-4 performs better than ChatGPT because it achieves higher approx. percentile scores.