But what do each of these three AI chatbots offer, and which is best for you?
ChatGPT
Claude
Perplexity
Availability
Worldwide (excl.
North Korea, Iran, China, Cuba, Syria, and Nepal).
Jason Montoya / How-To Geek
US, UK, Canada, Ireland, Australia, New Zealand.
LLM Used
GPT-3.5 and GPT-4.
Perplexity: GPT-3.5 Perplexity Pro: GPT-4, Claude
Price
GPT-3.5: Free.
Hannah Stryker / How-To Geek
GPT-4: $20/month.
Claude Pro: $20/month.
Perplexity: Free Perplexity Pro: $20/month | $200/year
Training Data
Up to September 2021.
Up to early 2023.
Perplexity: Up to September 2021 with GPT-3.5.
Perplexity Pro: varies depending on LLM used.
Token Limit
GPT-3.5: 4,096 tokens.
GPT-4: 8,192 tokens.
Varies depending on LLM used.
Now, let’s dive into the specifics of these three AI chatbot tools.
LLMs Used
AI chatbots rely on large language models (LLMs) to process and respond to prompts.
While ChatGPT and Claude use their own original LLMs, Perplexity offers a more varied choice.
ChatGPT usesGPT-3.5 and GPT-4.
If you’re using the free version of ChatGPT, you’ll be interacting with GPT-3.5.
Claude, created by Anthropic, uses its most recent LLM version, Claude 2.1.
Perplexity is a little different.
This is an AI chatbot platform that gives you access to multiple LLMs.
The free version of Perplexity lets you access GPT-3.5 in addition to its Copilot feature.
Be sure not to confuse this Copilot withMicrosoft’s AI chatbotthat goes by the same name.
On Perplexity’s platform, Copilot is a GPT-4-powered internet search tool that was launched in May 2023.
The Pro version gives you access to GPT-4 and Claude, too.
It’s important to note that there is no original Perplexity LLM.
The platform uses preexisting LLMs, not its own.
This premium version gives you access to OpenAI’s latest LLM version, GPT-4, for $20 monthly.
A Claude account is entirely free and gives you access to Anthropic’s latest LLM version, Claude 2.1.
Perplexity is available in most countries worldwide, though it is not available in the EU.
Therefore, the input and output limits for GPT-3.5 are both 2,048 tokens.
However, the output limit is only 2,000 tokens, which is worth keeping in mind.
However, Perplexity’s token limit for Copilot searches alone is 2,000 tokens.
Perplexity’s basic version also only allows five Copilot requests every four hours.
If you upgrade to Perplexity Pro, this shoots up to over 300 requests per day.
Since AI chatbots have entered the mainstream, there have been concerns about the accuracy of their responses.
This is where inaccuracy can pose an issue.
ChatGPT, Claude, and Perplexity all stand the chance of providing false information.
This isn’t intentional, and can come from biased or inaccurate resources used in training.
Therefore, the best thing to do is fact-check any information given by an LLM via another source.
ChatGPT has shown that it can hallucinate in the past, as discussed in anNCBI study.
Anthropic has also statedon its own websitethat Claude can hallucinate.
Conversational Abilities
LLMs are designed to interact with users in a human-like manner.
So, how do ChatGPT, Claude, and Perplexity measure up here?
Oftentimes, AI chatbots will provide you with bulleted lists if your prompts are quite loaded.
We tested this on ChatGPT-3.5 and Claude using the following prompt:
I need some advice.
Can you tell me the best way to approach an interview for a marketing assistant?
The interviewer is looking for concise yet informative answers, and I want to come across as confident.
These are the two responses we got from each chatbot.
Here are ChatGPT’s results:
ChatGPT gave a longer response with more bullet points and information.
The language was very conversational and natural.
Evidently, the GPT-3.5 responses differed on ChatGPT and Perplexity, which is important to note.
Some similar points were touched upon, but you certainly won’t be getting identical responses on both platforms.
This isn’t a simple equation, but a fairly difficult problem that students may come across.
Here’s Claude response:
Claude also gave the correct answer, along with the solution process.
Check out our piece onwhy you shouldn’t use ChatGPT for mathematicsto learn more.
In the end, selecting the right chatbot comes down to your budget, intended use, and location.
Consider the factors above if you’re trying to choose between ChatGPT, Claude, and Perplexity.
