Is Claude 3 better than OpenAI GPT-4?

Coaching Bubble App Builders

I coach and tutor many Bubble app builders each week and a lot of them are making really exciting AI powered apps. Most of them, in actual fact all of them that I've worked with so far, are using OpenAI's API to power the apps. They're tapping into the same power that powers ChatGPT and they're using mainly GPT 3.5 Turbo and GPT 4.

Exciting News: Claude 3 Outperforms GPT-4

Well, we've got a little bit of exciting news today. If you are using OpenAI, then take this as your encouragement to try out Claude 3. Based on this table where they claim they've collected over 400,000 human preferences, basically people get to vote on the output of different AI models and it's all gone through a statistical algorithm, we get a table like this and at the top what do we see? We see Claude 3 Opus.

Claude 3 Models: Opus, Summit, and Haiku

As a reminder, if we go back to Claude's press release, we can see that Opus is their most expensive and yet also their most powerful model. In terms of pricing, there's a slight difference but think of it like GPT 4. Here's where it gets interesting. As you scroll down, you can see we go past some GPT 4 previews, there's Bard, then we've got Summit which is their mid-range model, then we've got another GPT 4, and then we've got Haiku. Haiku is the baby of the Claude 3 models. It's so cheap and yet we're seeing here that it outpowers other GPT 4 models, more powerful than Claude 2.

Comparing Claude 3 to GPT Models

We have to scroll all the way down to the 14th place to get a GPT 3.5 turbo. Definitely don't quote me on the direct kind of token figures, but you should really do a price comparison and an assessment within your app of the difference between using GPT 3.5 turbo and using Claude 3 Haiku. At least according to these early results, remember that Claude 3 has only been out for a few days or weeks at this point, so you'll see that Claude has not got as many votes as some of the GPT 4 preview models.

Anthropic: A Serious Competitor to OpenAI

I've been saying this for a few months now: Anthropic are a serious competitor to OpenAI. If you have just been using the OpenAI playground to build your prompts and you've been using Bubble to build no-code apps with AI integrated into them, then yeah, this is your prompt to go and try out Claude 3. Try out each of the models available and see whether you believe that it outperforms OpenAI's GPT models.

Share Your Experience with Claude 3

I'd really love if you could leave a comment down below if you tried it and disagree with me, or disagree with the very early data that's come out about which is best. Please report your results in the comment section down below because I'd be really interested to know. Based on the data, I'm starting to recommend Claude to people. I'm saying, "Have you tried Claude 3?" Because lots of people know ChatGPT, they may have heard of OpenAI, but they probably haven't heard of Anthropic.

Try Claude 3 and Share Your Feedback

So here it is: go to Anthropic and try out the Claude 3 chatbot large language model. Let me know how your results are going and what sort of feedback you get. If you're learning Bubble and you're wanting to build no-code apps with AI, then follow the link down in the description because we've got hundreds of Bubble tutorial videos where we go really in-depth and some beginner content on how to build no-code apps with Bubble.io.