2023 Comparison: ChatGPT vs Claude 2 vs Bard – The Best Intelligent Chatbots

In today’s digital landscape, intelligent chatbots are redefining the way we interact with technology. These virtual assistants, powered by Artificial Intelligence, are changing communication, customer service, and many other aspects of our daily lives.

But how do they compare to each other? Which is the most efficient, the most intuitive, the smartest? In this article, I test the most prominent intelligent chatbots, comparing them to provide you with a clear view of which one might be the best for your needs. I invite you to a deep and detailed analysis in the universe of intelligent chatbots.

However, before we dive into the analysis, as I always prefer to do… let’s start from the beginning!

What are intelligent chatbots?

Intelligent chatbots are software designed to simulate human conversations thanks to the use of Artificial Intelligence. They use machine learning algorithms to understand and respond to user inputs, allowing them to interact with people in a more natural and conversational way. These chatbots are used in a variety of applications, from customer service to personal assistance and content generation.

The best intelligent chatbots of today

In the world of intelligent chatbots, there are some that stand out for their ability to interact effectively and naturally with users. Next, we will explore some of the best AI chatbots available in 2023, starting with ChatGPT.


ChatGPT, developed by OpenAI, is a language model that has proven to be extremely versatile. It is capable of generating coherent and relevant text in response to a variety of inputs, and has been used in a wide range of applications, from creative content generation to programming assistance. ChatGPT is free in its GPT-3.5 version, but to access the more powerful version that uses GPT-4, you will have to pay $20 per month for the ChatGPT Plus version.


Bard is Google’s chatbot designed to provide detailed and contextually relevant responses. Although Bard is free, it is currently available in fewer countries than ChatGPT. However, if you want to use it, you can ‘cheat’ and use a VPN to bypass its geographic zone restrictions. Bard has the option to auto-delete interactions in 18 months and does not allow users to retrieve previous interactions.

Claude 2

Claude 2 is the new language model from Anthropic. This chatbot has been designed to handle large amounts of text, up to 100,000 tokens, which is approximately equivalent to 75,000 words, in a single message. This allows it to provide responses in a broader context and more accurately. Claude 2 has also proven to be particularly effective in tasks related to programming and creative writing.

Comparison of features

Before we dive into the direct comparison, it’s important to understand that each of these chatbots has been designed with a unique set of features and capabilities. These differences can make one chatbot more suitable for certain tasks or applications than others. Below, we will compare these chatbots in terms of price, availability, privacy, context handling, and additional features.


ChatGPT is free for those who use the GPT-3.5 version. Those who want to use the more powerful version that uses GPT-4 will have to pay $20 per month for the ChatGPT Plus version. Both Claude 2 and Bard are free.


ChatGPT is the most widely available of the three. Bard is available in fewer countries than ChatGPT. Claude 2 is temporarily available in the UK and the US. But as I mentioned a little earlier, you can always cheat a bit to get around these restrictions. Bless VPNs!


ChatGPT allows users to delete their interactions but does not support browsing through VPN. Bard has an option to auto-delete interactions in 18 months and does not allow users to retrieve previous interactions. Claude 2 allows users to delete their conversations.

Context handling

ChatGPT supports 7096 tokens and up to 8,192 tokens of context in its Plus version. Bard supports 8,196 tokens of context. Claude 2 surpasses both with a context handling capacity of 100,000 tokens.

Supported languages

ChatGPT is the most versatile in terms of supported languages, with over 80 languages available. With the most updated Bard, you can collaborate with Bard in over 40 languages, including Arabic, Chinese, German, Hindi and Spanish. Claude 2 supports several widely spoken languages, such as English, Spanish, Portuguese, French, Mandarin, and German, among others. If Claude 2 does not recognize a language (or the input has many grammatical errors), it provides an introductory phrase and then responds in English.

General Features

Let’s delve into the characteristics that define these intelligent chatbots. We will analyze key aspects such as the available plugins, the user interface, the feedback options they offer, and the possibility of interacting with them through mobile applications.


In this section, we will explore the different extensions and additional functionalities that each chatbot offers through its plugins.

ChatGPTBardClaude 2
DescriptionPlugin store, code interpreter, web navigation function (temporarily paused) in Plus version.In experimental phase, plans to have a plugin store and integration with Google Suite.Can be added to Slack, summarize threads, provide suggestions, brainstorm.

In my humble opinion, I think OpenAI is ahead in this aspect, and that the chatGPT plugin store is the one that offers the best options to the user. You know, thanks to the plugins, these intelligent chatbots are a little bit more so, because they break barriers of connections with third parties or access to the Internet to obtain 100% updated information.

Communication with the chatbot

Here, we will analyze how each chatbot manages the interaction with the user, including file management, the possibility of voice interaction, and whether or not they have a web application.

File Management
ChatGPTBardClaude 2
DescriptionSimple interface, allows uploading a file using the Code Interpreter plugin.Allows attaching images in jpg, png, or webp formatSupports uploading files in pdf, txt, csv, etc. no more than 10Mb. Allows uploading up to 5 files.

ChatGPT Plus adds one more element to this interface, allowing the user to upload a file.

They do not make clear the type of files that can be processed, so you can upload any type, although when you try to ask the chat about it, you will find in its response whether it could process it or not.

You can see the full conversation here if you prefer.

Claude allows uploading files (in plural) in its interface, and in this case, it does show you what types of files it is capable of processing. For now, files in pdf, txt csv … and an etc that we don’t know exactly what it admits. But no more than 10Mb. Also, unlike chatGPT using the Code Interpreter plugin, Claude 2 allows you to upload up to 5 files.

Bard is 100% specific. It allows attaching files, but only images in jpg, png, or webp format. It does not indicate the maximum size, however.

What really sets Bard apart and makes it stand out is that it doesn’t just update text files like Claude 2. No, Bard ups the game with the ability to handle images. How? By incorporating Google Lens. With this, you can do such amazing things as giving it a sketch of your homepage and having it return the code, or simply helping you to describe any image. The only ‘but’ (for not English speakers) is that this feature is only available in English at the moment, although they promise to launch it in more languages soon.

Voice Interaction
ChatGPTBardClaude 2
DescriptionDoes not offer this functionality (in the web version).Admits voice interaction, can read responses aloud.Does not offer this functionality.

Bard admits that it can be spoken to directly using the microphone.

And not only that, but it also reads its responses.

It sounds a bit like a robot, but for being in Spanish, the truth is that it is very well achieved. It is a differentiating point (for now) of Bard, which makes it very comfortable to use.

Web Application
ChatGPTBardClaude 2
DescriptionOffers mobile app, allows speaking directly, transcribes messages in real time.Does not offer this functionality.Does not offer this functionality.

OpenAI launched the mobile app for chatGPT, something that neither Claude nor Bard offer to date. The chatGPT app offers certain advantages, such as being able to speak directly, and using Whisper, it transcribes your message to chatGPT in real time. For the moment the chatbot’s response is not read.

Another thing to note about the chatGPT mobile app is that when new conversations are started from its interface, plugins cannot be chosen. However, if you have other conversations created from the web, where you did have some plugin, then you will be able to make use of that or those plugins in the mobile app. A bit strange, really, but that’s the way it is.


While all 3 chatbots show in one way or another feedback options for their responses, Google seems to have been the one that has cared most about this, and its Bard chatbot seems the most complete in this aspect, offering the clear option to give quite a bit of feedback and report any type of concern.

It’s just my feeling, but I think Google is a bit ahead in this.

Visual Comparison

An excellent way to understand the differences between these chatbots is to see them in action. Below, we present three examples of how each chatbot responds to different types of requests. For each example, we will provide a screenshot and a brief description of the results.

Blog Idea Generation

We’re going to ask our 3 friends for help to see if they can suggest interesting and original topics to write about that are about Artificial Intelligence.

The prompt to use will be:

I'm planning to write a series of articles for my blog about artificial intelligence. Could you suggest some interesting and original topics that could attract my readers?


Even though in the screenshot you can see that the interaction is in Spanish, in this link to the complete conversation you can see that the last action was to translate the entire conversation into English, in case you want to take a look in your language. I will keep the rest of the screenshots on the page as they were originally in Spanish, but they will follow this same tone, so, if you want, you should be able to access to the whole chat clicking in the link.

ChatGPT suggested a series of topics that include AI in medicine, ethics in AI, AI in art and creativity, AI and sustainability, AI and sports, AI and cybersecurity, AI and education, AI in the financial sector, AI and agriculture, and AI and healthcare. The response from ChatGPT 3.5 is detailed and provides a description of each topic, demonstrating its ability to generate detailed and well-thought-out ideas. Plus, it was super fast.

ChatGPT Plus

Heres the link to the whole conversation.

ChatGPT Plus proposed topics such as ethics in AI, unusual applications of AI, AI and sustainability, the history of AI, how AI is changing the workplace, AI in health, the role of AI in combating misinformation, AI and education, the future of AI, and mitigating bias in AI. The response from ChatGPT 4 is similar to that of ChatGPT 3.5, but with some variations in the suggested topics. Perhaps where the difference is most noticeable is in the time it takes to generate the complete response, much slower than its predecessor.

Claude 2

Claude does not allow us to generate a Link to the conversation.

Claude 2 provided a detailed list of topics, which include ethics in the development and use of AI, creative and unexpected uses of AI, AI in developing countries, the impact of AI on different industries, AI and jobs of the future, interviews with AI leaders, explaining key AI concepts, and the history and philosophy of AI. The response from Claude 2 is quite complete and covers a wide range of topics, demonstrating its ability to generate creative and relevant ideas. It was quite fast in its response.


Link a la conversación.

Bard suggested topics such as the future of AI, the ethical challenges of AI, the use cases of AI, the applications of AI, the people working in AI, the history of AI, and the prospects of AI. Bard’s response is more general and less detailed than the other chatbots, but still provides a variety of interesting topics. Bard, unlike its buddies, doesn’t release the response until it’s fully completed. It was quite fast, but it generates a different feeling in the user. Personally, I like it less.

In general, all the chatbots provided a variety of interesting and relevant topics. However, Claude 2 and ChatGPT provided more detailed and creative responses, while Bard provided a more general response. This can be useful for readers looking for an overview of possible topics, but may not provide the same depth or detail as the responses from Claude 2 and ChatGPT.

Summary of a long text

Now we will try to evaluate the ability of the chatbots to manage long texts. The text I have chosen is from the book Don Quixote de la Mancha, whose full text is found in the Gutenberg project. For cases where the url does not work, I will use the content of chapters 1 to 3, both included.

The prompt to use will be:

Please summarize the following article for me: [insert the link of the article or the text of the article here]


When asked to summarize using the url

When doing so by including the text of chapters 1, 2, and 3 (I will not show a complete screenshot for obvious reasons), there was something that caught my attention and that I had not seen until now.

It seems that the chat was complaining because the prompt could infringe privacy policies… 🤔.

But then came the message I was expecting, although it is not the desired one…

And it is that chatGPT, as we commented, has a context window limited to 7096 tokens.

Here is the previous conversation. You will see that it only shows the first message. This is because chatGPT when it gives the error, it does not store that in the conversation. It waits for you to edit your message so that it can proceed successfully, but if you do not, that part of the conversation is not saved.

ChatGPT Plus

The result using the url in the prompt is the same.

And when I added the text of chapters 1, 2, and 3 I got the same result as with chatGPT.

Here is the link to the entire conversation. Just like with chatGPT, you will see nothing but the content of the screenshot above.

Claude 2

Here comes the first big difference. When we introduce the prompt to Claude to summarize the url… we have a surprise!

But not exactly a good one… Claude is hallucinating and is trying to ‘trick’ us by saying that this is the summary he considers appropriate. As you can see, it is something totally invented. Claude does not have the option to access the Internet, but he does not tell us, and tries to give an answer… even if it is invented. Negative mini-point.

We have certain tricks that we can apply to try to avoid this. I talked about it in this article about GPT. But regardless of whether we can reduce hallucinations or not, the response is a clear indication of Claude’s ease in always responding, regardless of whether it’s a hallucination.

When I record in a txt the content of chapters I, II, and III and upload it, it is able to give a summary. First it gives it in English, I understand that because the content of the book I have uploaded is in English, but then it is asked in Spanish and it does a good job.

This is where Claude’s strong point is seen, with his ability to handle long texts.


Bard acts the same as Claude when we indicate that he should give us the summary of the url.

This issue of hallucinations is the worst thing about these LLMs, because they can make something that is not at all credible seem credible. As you can see, it again invents the answer as best it sees fit instead of simply telling us that it cannot give an answer because it does not have access to the Internet. By the way, Google earned that same feedback from me when it gave me this answer!

When we move to the alternative option, of giving him the text of the 3 chapters, Bard has a different behavior.

It directly does not let the user introduce more than the number of characters it stipulates, cutting the entry up to that point. Bad, because the user may not notice it.

In addition to that, I gave it to execute to see what it did and … this was its answer:

Here is the link to the entire conversation.

In general, all the chatbots showed different capabilities and limitations when handling long texts.

ChatGPT and ChatGPT Plus, despite their versatility, encountered limitations due to their 7096 token context window, which prevented them from processing and summarizing the full text of the three chapters of Don Quixote.

Claude 2, on the other hand, demonstrated its strength in handling long texts by providing a coherent and detailed summary of the provided text. However, its inability to access the Internet became evident when it tried to summarize the content of the URL, providing an invented response instead of admitting its limitation. This is very dangerous.

Bard also showed similar limitations to those of Claude 2. Although it could not summarize the content of the URL, it opted for an invented response. It also demonstrated a limitation in the length of text it can process, cutting the user’s entry after a certain number of characters. And even so, being unable to give a half-coherent response in that case.

In summary, while Claude 2 shines in handling long texts, its lying behavior in the face of lack of Internet access can be a major downside to consider. On the other hand, both ChatGPT and Bard need improvements in their ability to handle long texts, expanding their context windows.

Assistance in programming

Let’s see which chatbot would be a better co-pilot when it comes to software development!

First exercise

The prompt to use will be:

I am trying to write a function in Python that takes a list of numbers as input and returns a new list that contains only the even numbers from the original list.

Could you help me write this function?


Here is the entire conversation.

ChatGPT is like your favorite math teacher. It explains how to make a function in Python that filters the even numbers from a list. It tells you to use a loop, which is like a process that repeats, to check each number. If the number can be divided by 2 without leaving anything (that’s what the % operator does), then it is even and you add it to your new list. In the end, you are left with a list of only even numbers. It also shows you how you can test your new function with a sample list.

ChatGPT Plus

Here is the entire conversation.

This is the cool kid in the class who knows all the tricks. Instead of using a loop, it uses something called “list comprehension” to do everything in one line. It’s like doing a magic trick: fast and elegant, but you might need a little more practice to understand it.

Claude 2

Imagine Claude 2 as a chef teaching you how to make a recipe step by step. It tells you to take a list of numbers and use a loop to check each number. If the number is even, you add it to your new list. In the end, you are left with a list of only even numbers. Very similar to chatGPT.


Here is the link to the entire conversation.

Bard is like your trusted friend who is always willing to help you with homework. It gives you an explanation similar to those of Chat GPT 3.5 and Claude 2, and also shows you how to use the function with an example.

In summary, all the bots provided correct and useful answers, but with different levels of detail and different coding styles. Depending on your level of experience with Python and your personal preferences, you may find one answer more useful or easier to understand than the others.

Second exercise

I was not completely satisfied with that first exercise and decided to try adding a little bit of difficulty to the matter.

This is going to be the prompt we use now:

I am trying to write a function in Python that takes as input a long text, which can be a book or an article. The function should divide the text into words and count the frequency of each word. Then, it should return the 10 most frequent words and their number of appearances.

In addition, I want the function to ignore common words in English such as 'the', 'a', 'an', 'in', 'on', etc. I also want the function to be able to handle different forms of the same word such as singular, plural, and different verb tenses like 'run', 'ran', 'running'.

Finally, the function should be able to handle text in different languages, not just English. Could you help me write this function?


Here is the link to the entire conversation.

It shows you how to write a function in Python that uses the NLTK library to split the text into words, filter common English words, lemmatize the words and count their frequency. Then, it sorts the words by frequency and returns the 10 most frequent words. It warns you that you need to have the NLTK library and additional resources downloaded for the function to work correctly. It feels like a good teacher explaining things to you.

ChatGPT Plus

Here is the complete conversation.

Again it shines in its response. It seems like an expert in natural language processing. It provides you with a code similar to that of Chat GPT 3.5, but warns you that the task you are describing is quite complex. It explains to you that handling different forms of the same word is called “lemmatization”, and that handling text in different languages can be quite difficult depending on the language.

It also points out to you that this code will only work for English text and that to support other languages you would need stopwords and a lemmatizer that works for those languages. Very top.

Claude 2

Continuing with the comparisons, we could say that Claude 2 is like an experienced software engineer. It gives you a code that first cleans and tokenizes the text. Then, it removes common words and lemmatizes the remaining words to group different forms.

Finally, it uses Counter from collections to count the frequencies and return the most common words. It tells you that the function is generic to handle different languages and can be easily modified to ignore specific words of a language.


Here is the complete conversation.

It returns a code that converts the text to lowercase, removes common words, splits the text into words, creates a dictionary to count the frequency of each word, sorts the words by frequency and returns the 10 most frequent words. It tells you that this function can be used to analyze text in any language and that you just need to change the list of common words to include the common words in that language.

In summary, each bot has its own style and approach to solving the problem, resulting in solutions with different strengths and weaknesses.

Chat GPT 3.5 and Chat GPT 4, for example, offer robust solutions that include lemmatization, allowing to handle different forms of the same word. However, these solutions depend on the NLTK library, which could be a drawback if you are working in an environment with installation restrictions.

On the other hand, Claude 2 offers a solution that does not depend on any external library, which is a great advantage if you cannot install NLTK. But this solution may not be as effective in handling different forms of the same word, as it does not include lemmatization.

Bard, finally, offers a solution that seems to be a compromise

between the others: it includes lemmatization, but it is not clear how it would handle different languages.

Therefore, the best answer for you will depend on your specific needs and the limitations of your environment.


Another very used function in these chat bots is to translate. Let’s see how they behave in this task!

The prompt to use will be:

I'm working on a novel and I have a paragraph that I would like to translate from English to French to include in the French version of my book.

Here is the paragraph: 'In the quiet town of Veridian, where the trees were tall and the air was filled with the scent of blooming flowers, lived a young girl named Amelia. Amelia was not like the other children. She had a vivid imagination and spent her days exploring the vast forest, creating magical worlds and befriending mythical creatures.'

Could you help me translate this paragraph into French?


Here is the complete conversation.

In an almost instantaneous response, ChatGPT says that… that’s the translation. Honestly… I have no idea if it’s right or not! So… I used Google Translator to verify.

And apparently… the result in English is quite similar to the input of our original message. So… good!

ChatGPT Plus

Here is the complete conversation.

The translation is very similar to the one provided by Chat GPT 3.5, indicating that both bots have a good handle on translating from English to French.

Claude 2

Since we had the original chat in Spanish, the chatbot is answering like that when I asked it to translate. The original translation to French was this:

Claude 2 provides a translation that is practically identical to those provided by Chat GPT 3.5 and Chat GPT 4. This shows that Claude 2 is also capable of making accurate translations from English to French.


Here is the complete conversation.

Bard offers a translation that, while slightly different in terms of word choice, is still accurate and maintains the tone and style of the original. This shows that Bard is also capable of making accurate translations from English to French.

In summary, all bots demonstrated the ability to make accurate translations from English to French. Although there were small differences in the choice of certain words, all translations maintained the tone and style of the original. The ability of these bots to handle complex translation tasks is evident.


Let’s see how the bots behave with a prompt that requires a bit of creativity from them.

The prompt to use will be:

I'm working on a collection of poems for children and I would like to include a poem that teaches the importance of friendship. The poem should be set in a magical world and should include a friendly dragon, a brave child, and a lost treasure.

Could you help me write this poem?


Here is the link to the chat.

ChatGPT created a poem that not only meets the requirements, but is also quite lyrical and poetic. The story is well developed and the importance of friendship is effectively highlighted. The poem has a nice rhythm and uses descriptive language to paint a vivid picture of the magical world.

ChatGPT Plus

Here is the complete conversation.

ChatGPT Plus also created a pretty cool poem that follows the story of a boy and a dragon in search of a treasure. The poem has a consistent rhythm and rhyme, making it enjoyable to read. The friendship between the boy and the dragon is highlighted throughout the poem, and the revelation that true friendship is the real treasure is a powerful message.

Claude 2

Claude 2 provided a poem that is a bit simpler in its structure and rhyme, but still effective in its message. The story follows a dragon named Timby and his new friend Tom, and how their friendship helps them overcome challenges and find a treasure. The poem has a steady rhythm and a clear rhyme structure, making it easy to follow for younger readers.


Here is the link to the complete conversation.

Bard provided a poem that is quite simple and straightforward. The story follows a boy named Juan and his friendship with a dragon. Together, they find a treasure and decide to share it. The poem is quite short and the rhyme is simple, which could be suitable for younger readers. However, the story is not as detailed or descriptive as those provided by the other chatbots.

In summary, all chatbots were able to create a poem that met the requirements of the request. However, the responses from ChatGPT 3.5 and ChatGPT 4 were quite more detailed and poetic, which could make them more appealing to readers looking for a richer and more descriptive story.

Claude 2 and Bard provided simpler poems that might be more suitable for younger readers or those who prefer a clearer and more direct rhyme structure.

Up-to-date Information Responses

Let’s see how the bots behave with a prompt that requires them to be a bit creative.

The prompt to use will be:

I'm writing an article about the global economy and I would like to include the most recent information about the value of gold. Could you provide me with the current price of gold per ounce in US dollars?


Here the complete conversation link.

ChatGPT was clear in explaining that, as a language model, it does not have real-time access to updated information such as gold prices. It recommended consulting reliable online sources, such as specialized finance websites or financial news agencies, to get the current price of gold per ounce in US dollars.

ChatGPT Plus

Here’s the complete conversation link.

ChatGPT Plus, like ChatGPT, also explained that it cannot provide real-time information or live updates. It suggested consulting a reliable online source to get the most recent price of gold per ounce in US dollars, and provided examples of good places to look, such as financial news sites, commodities exchange websites, or pages specialized in precious metals.

Claude 2

Claude 2 also clarified that it does not have access to real-time data on current gold prices. However, it provided general information about the recent trend of the gold price, including its increase in the last decade, its historical high in 2020 due to the COVID-19 pandemic, and its price in mid-2022 due to high inflation and geopolitical uncertainty from the war in Ukraine.


Here’s the complete chat link.

Unlike the other chatbots, Bard provided a specific price for gold per ounce in US dollars. Looking it up on the Internet, I was able to verify that it is not the exact price, although it comes close.

In summary, all the chatbots provided useful responses, but none could provide real-time information. Bard was the only one that provided a specific price, although its accuracy, as checked, cannot be guaranteed. It seems Bard can leverage Google Search so that it manages up-to-date information.

On the other hand, Claude 2 provided general information about recent gold price trends, which may be useful for providing context in your article.


The choice of the ideal AI chatbot largely depends on the user’s specific needs and the context of use. Each of the chatbots we’ve examined, ChatGPT, ChatGPT Plus, Claude 2, and Bard, has its own unique strengths and weaknesses.

ChatGPT and ChatGPT Plus stand out for their creativity and ability to generate detailed and well-thought-out ideas. They are excellent for tasks that require specific language support, such as generating ideas for a blog or writing poems. Also for programming tasks.

However, they do run into limitations when handling long texts due to their context window of 7096 tokens. Although soon models will be released that already allow 32K tokens. In addition, the use of plugins gives them a point of versatility, with a connection to the outside world, which gives them a lot of potential. From my point of view, they are still a step ahead.

Claude 2, on the other hand, shines in handling long texts, providing coherent and detailed summaries of long texts. It is also capable of performing accurate translations and providing useful programming responses. However, its inability to access the Internet became evident when it tried to summarize the content of a URL, providing an invented response instead of admitting its limitation. This can be a point of concern, especially in contexts where the accuracy of information is critical.

Bard, developed by Google, shows great strength in handling tasks that require accuracy and facts. It leverages its Google Search connectivity to provide updated information, although its accuracy cannot be guaranteed. However, it may not be the best choice for creative tasks, as its responses tend to be more general and less detailed than those of the other chatbots.

Ultimately, you don’t have to choose a single chatbot. You can use all of them depending on your specific needs. For example, you can use ChatGPT or ChatGPT Plus for creative tasks, Claude 2 for handling long texts, and Bard for tasks that require updated information. At the end of the day, the choice of the ideal chatbot will depend on your needs, preferences, and the specific context in which you plan to use the chatbot.

Remember, learning is a journey, not a destination. So keep exploring, keep asking, keep learning. And while you do, don’t forget to enjoy the journey. See you in the next post, where we will continue to explore together the wonders and challenges of Technology. Until next time!