Debate: Claude-3 Opus vs. OpenAI GPT-4

General
,
1

I’m curious to hear community feedback on Claude-3 Opus performance vs. OpenAI’s models and am creating this thread as a space for customers to write in with their own thoughts. We’ve just added Claude-3 Opus to our no-code GPT builder.

Claude Advantages:

  • I’ve found that Claude is much better roleplaying and will add the asterisked roleplay stage directions without prompting.

  • Claude seems to have less annoying safety rails than OpenAI and is overall less likely to tell me “As an AI assistant I am unable to do X…”

  • Claude also has a 200,000 token context window.

Overall I am liking Claude-3 Opus and may be using it over OpenAI models.

2 Likes
2

I believe I was one of the first users to try this and while there were some preliminary bugs I feel very hopeful about the use of Claude. Previously I have not run into a way to create a personalized Claude model on other AI websites including Claude’s (in contrast to personas/GPTs) which has been very fun to play around with. I completely agree with your statement about fewer “safety rails” and think that Claude lacks a lot of the fluff seen in GPT models which I like a lot. I am thinking about switching all of my Pickaxes over to this model though I am still having some internal debate. If others have any opinions that may oppose this switch I would like to hear all of your opinions.

2 Likes
3

I’ve tested both Claude-3 and GPT-4 with a RAG system and find that GPT-4 outperforms Claude-3 in terms of detail and structure of the response. It provides well-organized summaries that include specific details and additional context at times. On the other hand, Claude-3 provides a more concise response (without lacking any detail based on the user input) and, in my opinion, is better suited for a chatbot environment.

1 Like
4

I use AI every day to perform complex business analysis tasks for my software development agency. I can tell you from my experience Claude 3 opus has a much better understanding of complex topics than GPT-4o. It also produces better writing and more useful breakdowns. GPT-4o seems to spam you with bullet point lists even when they’re not useful, whereas Claude understands when and how to break something down for you.

2 Likes