ChatGPT, GPT-4, Claude, Google Bard, Bing AI. If you feel invaded by chatbots, you're not alone. But beyond the initial saturation, the real problem is having to deal with the bureaucracy of their developers, creating multiple accounts and profiles. The alternative is to take a shortcut like OpenPlayground, a website that lets you test GPT-4 and dozens of additional models under the same interface, for free.
OpenAI requires an account and a phone number to access ChatGPT. If you want to use the new model GPT-4, you'll need to pay for a premium account. Bing's chatbot uses the same technology, but you can't talk to it without a Microsoft account (the Edge requirement disappears by using this extension). Google Bard is only available to users in the UK and US at the time of writing. And Anthropic Claude demands an “access request” that essentially works as a waiting list.
In other words, all the companies have thrown themselves into the waters of artificial intelligence, but the end user doesn't have such a clear path. Fortunately, there are platforms like OpenPlayground, which with a single account lets you test GPT-4, ChatGPT, Claude, and other secondary models, completely free.
How to Test GPT-4 and Other AI Models with OpenPlayground
The only requirement of OpenPlayground is creating an account (or using Google credentials) and a mobile number to verify entry. Once inside, you'll find a basic interface with just a column of advanced parameters on the right. Changing the model is as simple as selecting it from the list, and queries must be confirmed by clicking the Submit button.
However, that's not all: the Compare section is excellent for comparing responses from multiple models at once by loading a single prompt. Within that comparison, you can evaluate the time each model took, the speed in characters per second, and the final character count. OpenPlayground selects three models, but you can add more by checking their boxes.
Personally, I think model comparison is the most attractive feature of OpenPlayground. It helps us detect errors, limitations, contradictions, and moments of “delirium”. Finally, I imagine some users will dislike the idea of sharing a mobile number to confirm entry, but having one account per model is even more annoying. Give it a try!
UPDATE: The service is no longer free. The developer explains the reasons as soon as you enter: "It's been fun offering nat.dev for free, but some of you keep trying to bankrupt me, so please pay to cover inference costs. This way I keep losing money, but hopefully a finite amount of money." It was nice while it lasted.
Official site: Click here
