Categories
Tutorials

ChatGPT vs Claude 3.5 Sonnet

Swept away by Claude 3.5 Sonnet’s Artifacts, I further investigated whether Anthrophic’s alternative to OpenAI’s ChatGPT is something I should consider.

Context Window

Claude has a bigger context window of 200K tokens versus ChatGPT 4o’s 32K – 128K context window, depending on how you access it. On the API, you get 128K, but through ChatGPT, you might get anywhere between 32K-128K.

The implication for having a small context window is you’d notice your AI forgetting stuff. You won’t be able to analyze larger files because the content won’t be included in the analysis.

The only way to counteract that is to be more efficient in your prompting. So, avoid pasting parts that are not relevant to the answer.

In other words, the smaller the context window, the more work you’ll do for your AI.

Winner: Claude 3.5 Sonnet

Visualizing Projects in Artifacts

claude 3.5 sonnet artifacts generates svg

Holy moly, Claude’s Artifacts is a game changer.

ChatGPT has no equivalent, and frankly, Artifacts is a HUGE pull for me to use Claude.

Sometimes, you need to whip up a calculator or an SVG quickly. Artifacts cuts down the steps to visualizing your project.

Sure, ChatGPT will be able to generate the code, but Claude goes a step further and shows you how the app would look. Or how your vector image would look.

I wrote a more in-depth review here. This is the biggest selling point of Claude, IMHO.

Winner: Claude 3.5 Sonnet

Claude’s Projects vs ChatGPT’s Custom GPT

Way back, you’d have to save your prompts into a txt file and then paste it into your prompt window. Not these days.

Claude’s Projects feature allows users to create a custom work environment with shared information among different chats. ChatGPT offers this flexibility with custom GPTs.

The difference isn’t huge between the two. The main thing is that Claude just included Projects in late June, whereas custom GPTs have been around for a much longer time.

That means if I wrote this article just a month ago, I would be saying Claude has no comparable functionality…

Winner: Tie

Browsing Capability

ChatGPT can go online (sometimes) to find stuff for you.

Claude can’t.

But I try to avoid ChatGPT’s online browsing. I prefer to just take a full-page screenshot or copy and paste the text into the prompt.

Why? Because my experience with ChatGPT’s browsing is poor. It can sometimes only analyze a portion of the content. It can sometimes fail to retrieve the website.

It’s not reliable.

So anyway, it’s a win for ChatGPT but not a big win.

Winner: ChatGPT (but barely)

Content Censorship

Claude 3.5 is more wary about what you ask it to do. ChatGPT is more libertarian.

Some users find it more restrictive in terms of content moderation. So if you have mention of gore, or anything that wouldn’t fly in Teletubbies, Claude is more likely to shut it down.

I do create content about dentistry and health care sometimes. This content gets checked by a dentist or doctor, so to me, there’s little risk in hallucinations causing dangerous advice. Therefore, the best LLM is one that gives me the answer without filtration. I’ll leave the filtration to the experts in their domain.

Winner: ChatGPT

User Experience

Some have said they prefer Claude’s user interface. I suppose with their serif font and brown background, it feels more like writing into a magic book.

The grey-white colour scheme of ChatGPT though… feels like an Apple colour scheme.

To me, there’s not much of a debate here because it’s an each to their own preference. I am more familiar with ChatGPT, therefore its interface works better for me.

Winner: You (because more variety is good for the consumer)

Rate Limits

I have been hearing that Claude users are more likely to hit rate limits.

Here’s Anthropic has this policy for its subscribers:

  • At least 5x the usage compared to the free service.
  • Message limit resets every 5 hours.
  • Expect at least 45 messages every 5 hours if conversations are relatively short.

Comparatively, I can’t remember the last time I hit a rate limit on ChatGPT Plus. In fact, ChatGPT has completely done away with their fine-print warning that “you have 40 messages in 3 hours”. A few months ago, you’d still see it.

I don’t want to make a conclusion here because I don’t have first-hand experience, but I’d be a bit cautious about using an LLM which has strict rate limits.

Think of it this way: when ChatGPT was down, people couldn’t work.

In those two years since release, it has certainly delved into every aspect of my workflow.

Winner: ChatGPT

Cost

They’re the same at US$20.

So, just judging from a cost perspective, I’d go with ChatGPT.

Of course, if you are an archaeologist and love Artifacts, then it might make sense to get Claude Pro because iterating on an app can take many prompts.

Winner: ChatGPT

Have a project in mind?

Websites. Graphics. SEO-oriented content.

I can get your next project off the ground.

See how I have helped my clients.