ChatGPT vs Claude 2.1 - Side by Side Comparison (2024)

콘텐츠

ChatGPT has a new competitor – Claude 2.1, and in this testing of ChatGPT vs Claude, you will see how good this new competitor is.

Anthropic, a company formed by former OpenAI employees, is built to be helpful, harmless, and honest. It brings with it the ability to accept a large number of tokens, generate faster responses, and provide detailed, unbiased answers.

ChatGPT, on the other hand, has seen massive improvements with the new features and updates. It will be an exciting duel when compared to some of the common uses for which users use an AI chatbot.

Let me take you through this comprehensive comparison of ChatGPT and Claude and show you which one is worth your money!

ChatGPT vs Claude 2.1: At A Glance

Here’s a glance at the comparison between ChatGPT and Claude 2.1. If you are in a hurry or are on the go and don’t have enough time, the below section will serve your curiosity.

Note: I have compared Claude 2.1 with ChatGPT Plus in this article.

| Parameters | ChatGPT | Claude 2.1 | | -------------------------------------------- | ---------------------------------------------- | -------------------------------------------------------------- | | Parent Company | OpenAI | Anthropic | | Financially backed by | Microsoft | Google | | Launched on | 30th November 2022 | 14th March 2023 | | Number of Users | 100 million+ | Available Only For Beta Test | | Head Quarters | San Francisco | Amsterdam | | Founders | Sam Altman(CEO), Elon Musk(Co-founder OpenAI), | Daniela and Dario Amodei | | Availability | Available in more than 159 countries | Available in over 80 countries | | Technology used | GPT-3.5 and GPT-4 | Constitutional AI | | Price | Free and ChatGPT Plus at $20/month | Free and Claude PRO at $20/month | | Token Limit | 8,192 | 200,000 | | Web browsing | Available with plugins | Not Available | | Visual Abilities | Can process and interpret images | Does not accept image inputs. | | Data Analysis | Easily analyzes huge datasets | Can analyze large datasets (up to 150,000 words) | | Calculations. | Better at explaining complex calculations | Gives basic explanation for calculations | | Factual understanding and logical reasoning. | Better than Claude | Incorrectly answers factual queries. Good at logical reasoning | | Creativity | 4/5 | 3.8/5 | | Code generation | Provides basic codes | Provides detailed and simplified code. | | Humor | Same as Claude 2 | Not as humorous | | Text summarization | Well summarized. | GPT-3.5 and GPT-4 |

Let us now look at the detailed features of ChatGPT and Claude 2.1 chatbot.

ChatGPT vs Claude 2.1: Overview

Let us take a quick sneak peek at ChatGPT and Claude 2.1 and what they have to provide the users before making a detailed comparison.

ChatGPT Overview

ChatGPT Plus has been proving its abilities since its launch, and again, with the addition of plugins and a Code interpreter, it has become unstoppable.

It has further seen more improvements in terms of its ability to understand a user prompt via file, audio, or image than simple texts. The GPT 4 subscription opens newer doors for a user to use ChatGPT.

Demandsage Bonus: To get more insights into the facts and figures related to ChatGPT, you can check out the latest ChatGPT statistics.

What’s New With ChatGPT

There have been a ton of updates to ChatGPT in a few months. As I mentioned, text is not the only way to interact with ChatGPT now. Check out all the major new updates to ChatGPT:

  • Mobile application for Apple and Android users.
  • AI image generation with DALL E.
  • Set “Custom Instructions” for better results.
  • Have a back-and-forth voice conversation with an assistant.
  • Image-to-text output.
  • Internet access with “Bing Search.”
  • Access external plugins.
  • GPTs for specific tasks.
  • GPT store to choose from GPTs created by others
  • A team plan for sharing GPT access across the workspace, training data exclusion, and more!

Creating and using a custom GPT is super-easy; I used and explained this in the detailed guide, where I covered ChatGPT Chatbot and its recent updates, plugins, and use cases.

Claude 2.1 Overview

Antropic is again in discussions with the launch of Claude 2.1. Its improved content generation, coding, and other improved abilities have impressed the audience.

However, it is still in the testing phase and will soon be available in other countries as well.

What’s New With Claude

Claude launched its newest advancement to its language model with the introduction of Claude 2.1 back in November 2023. This new language model has double the amount of toke it can handle!

While Claude hasn’t seen updates in terms of new features such as ChatGPT, it has seen improvements to its already existing capabilities.

Check out what else is new with Claude:

  • Launch of Claude 2.1 with a 200,000 tokens usage limit.
  • Improvements in comprehension and summarization (for legal documents, financial reports, etc.)
  • 30% reduction in giving incorrect answers.
  • 3-4x rate of mistakenly concluding documents.
  • API tool use (in development).
  • Improvements to developer experience.

ChatGPT vs Claude 2.1: What’s the difference?

ChatGPT currently uses multimodal GPT-4 technology, which accepts text, image voice inputs.

Hence, ChatGPT became more powerful with recent updations like the integration of GPT-4, code interpreter, Plugins. It can generate creative content, crack jokes, Analyze data, process and generate images, etc.

On the other hand, Anthropic uses the method of constitutional AI to generate the output on Claude 2.1. This model provides the best answers to the users with the help of underlying principles called the constitution.

ChatGPT and Claude use reinforcement learning techniques to train the model and generate the required outputs according to the queries.

Earlier, the difference was small, but with the latest updates, ChatGPT seems to be way ahead of Claude. It is also older in comparison to Claude and offers a better experience in general.

Let’s take it step by step and break down the differences between token sizes, prompt processing times, knowledge base, etc.

**Token Size

A token generally refers to ¾ words, and a paragraph of an English text can contain anywhere up to 100 tokens.

You can see a significant difference between the number of tokens that both ChatGPT and Claude can process. Let’s first have a look at ChatGPT.

1. ChatGPT token size:

ChatGPT has different token limits depending on the model that you are using. Here are the token limits that you can use one time.

  • GPT 3.5 token usage limit– 4,096 tokens, translating to approximately 1500-2000 words.
  • GPT 4 token usage limit– 8,192 tokens, translating to approximately 3500-4000 words.

2. Claude token size:

Claude, on the other hand, works in a different league altogether when it comes to the number of tokens that can be processed. Earlier, when it was using the “Claude 2” language, the token limit was 100,000. But with the recent update, that’s gone even further.

  • Claude 2.1 token usage limit: 200,000 tokens that translate to approximately 150,000 words.

This means that Claude 2.1 can operate a number of tokens that are roughly 2300% higher than what ChatGPT can.

**Processing Time

A general rule of thumb is that the greater the number of tokens in a prompt, the more time is required to procure and generate a result.

The average processing time required to process a query on ChatGPT is 30 seconds. This time may vary according to the complexity of the question.

Did you know? The processing time limit for ChatGPT is officially 60 seconds, within which it should understand, process, code, execute, and respond to a query. Post this, ChatGPT times out and tries another approach.

On the other hand, I experienced the average processing time required by Claude 2.1 is 5 seconds, and this time varies if the query asked involves complex calculations.

**Knowledge base

Regarding the knowledge base of both these AI assistants, there is a slight debate over the knowledge cutoff.

While the official Open AI page shows a knowledge cutoff till 2021, ChatGPT insists that its knowledge cutoff has been increased to January 2022 but is unable to provide answers to events that happened in January 2022.

In comparison, Antropic claimed that Claude was trained on vast data and included information about the latest events. Besides, it improves itself with the help of the data received from the users.

What is interesting here is that when I asked both – “Who won the El Clasico held on 13 Jan 2022,” ChatGPT wasn’t aware of this information.

But Claude answered this question, while it also directly says that it cannot access the internet. Upon asking, “I don’t have a boundless knowledge limit,” Claude responded.

How did it answer this? I am still wondering!

**Availability and Supported Languages

ChatGPT is available in more than 160 countries and supports English mainly. Recent updates to ChatGPT have enabled browser language support for Chinese, French, German, Italian, Japanese, Portuguese, Russian, and Spanish.

While presently, Claude supports only the English language, it knows more than a dozen languages that it can translate. Apart from English, it is particularly good at Portuguese, French, and German.

**Pricing Comparison

ChatGPT has been free ever since its launch. But for using some advanced features like image generation, GPT usage, creation, plugins, etc., it has a separate pricing plan that uses GPT 4. Here’s how you can access GPT 4 on ChatGPT.

Here’s how the pricing for ChatGPT is lined out:

| Subscription type | Price | Features | | ----------------- | --------- | ---------------------------------------------------------------------------------------------------------------- | | ChatGPT | Free | GPT 3.5 | | ChatGPT Plus | $20/month | GPT 4, DALL E image generation, GPT store access, plugins, Bing search | | ChatGPT Team | $25/month | Ability to share GPTs with the team, increase message caps, advanced data analysis, and data training exemption. |

Claude 2.1 is free to use but has limited usage per day. However, there is no clear figure mentioned for the free daily usage. While using Claude, a message suddenly popped up highlighting the number of messages I could generate till a certain time.

Here’s an answer from the official FAQ page,

Yes. While using the claude.ai free open beta, there is a daily message limit. The number of messages you can send will vary based on demand, and Claude will notify you when you’ve reached your daily limit. Every morning, your quota will reset.“

Claude also offers a paid subscription with access to over 5 times more usage!

| Subscription type | Price | Features | | ----------------- | --------- | ---------------------------------------------------------------------------------------------------------- | | Claude Pro | $20/month | Access to 5 times more Claude usage, priority access, version switching, and early access to new features. |

Now, let’s compare and check the features that differentiate ChatGPT and Claude 2.1. I will be comparing both in terms of their ability to understand text and images, language, analyze data, creativity, logical reasoning, and accuracy.

Let’s get to it!

1. Text Summarization

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👎 | 👍 |

Summarizing long texts is where both ChatGPT and Claude 2.1 do well. Yet, in terms of understanding the number of tokens, Claude 2.1 is way ahead.

Where ChatGPT will stop processing text and point out that the text is too long, Claude 2.1 goes on!

I used our existing article mentioned above, “ChatGPT Chatbot,” which is a 5000+ word article, and asked ChatGPT and Claude 2.1 to summarize the content.

Here are the responses generated by Claude 2.1 and ChatGPT.

**Claude 2.1’s Response:

Claude 2.1 Text Summarization

Upon pasting, Claude 2.1 converted the content into a “.txt” file to understand it better. Towards the right, I was able to view the content it gathered to review if everything was accurate.

Claude 2.1 was lightning-fast to get started. It summarized the content well within 200 words and was able to explain the chunk of content as an overview.

ChatGPT’s Response:

ChatGPT Text Summarization

GPT 3.5 couldn’t generate a summary, but GPT 4, even without a plugin, provided a detailed summary when I copied and pasted the content.

When I used the “Webpilot plugin,” It saved me the trouble of copying and pasting the content. Since it can access the internet, I could simply take the URL of the article and paste it as a prompt.

Unlike Claude 2.1, ChatGPT took some time to generate a response. It wasn’t an eternity, but it was surely more. This might also be because of my using the plugin, but both ways I used it, I got an elaborate summary bifurcated point by point!

2. Image Recognition

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👍 | 👍 |

For the next comparison, it was already a win for ChatGPT as it can understand an image as a prompt. I used our post of 17+ AI image generators and took an AI-generated image to test if ChatGPT could recognize it.

Here’s ChatGPT’s response:

ChatGPT got the idea right and accurately decoded the image with a description of the subject, the setting, and the style used.

Claude 2.1, without its ability to understand an image as a prompt, could not generate a result.

3. Language Understanding

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👍👍 | 👍👍 |

Claude and ChatGPT are great at understanding foreign languages and providing explanations or translations of any language. I asked Claude 2.1 and ChatGPT to translate two idioms, one in Mandarin and the other in Sanskrit.

  1. Mandarin Idiom and its meaning: ‘心旷神怡’ – to feel relaxed and happy
  2. Sanskrit Idiom and its meaning: जलबिन्दुनिपातेन क्रमशः पूर्यते घटः = Drops fill up a pitcher

Let’s see how both of them did to understand the languages that are particularly not their forte:

Claude 2.1’s understanding of the idioms:

  1. Mandarin Idiom

“心旷神怡” (xīn kuàng shén yí) is an idiom in Chinese that means “carefree and happy”, or “delighted”.

It is used to describe a state of feeling relaxed, spacious, content, and joyful. A direct translation of the four characters is:

心 (xīn) – heart, mind

旷 (kuàng) – spacious, broad

神 (shén) – spirit, expression, look

怡 (yí) – joy, delight

So together, they evoke an image of one’s heart/mind being broadened and uplifted, resulting in a delighted spirit. It’s a very visual and poetic way to convey the feeling of happiness and contentment in Chinese culture.

  1. Sanskrit Idiom:

“Unfortunately, I do not speak Hindi and cannot provide a direct translation of the phrase you provided. However, here is an attempt at translating it based on breaking it down word-by-word using online resources:

जलबिन्दुनिपातेन (jalabindunipātena) – by the falling of drops of water

क्रमशः (kramashaḥ) – gradually, in sequence

पूर्यते (pūryate) – is filled

घटः (ghaṭaḥ) – a pot

Putting it together, a tentative translation would be:

“A pot is gradually filled by the falling of drops of water.”

It seems to convey the idea that something is filled incrementally by small, sequential additions of some liquid, quite likely water being used to fill a pot.

Please correct me if I have mistranslated or misunderstood any part of the phrase’s meaning! Let me know if you need any clarification.”

ChatGPT’s understanding of the idioms:

  1. Mandarin Idiom

“The Chinese phrase “心旷神怡” (xīn kuàng shén yí) translates to a state where the heart is broadened and the spirit is delighted. It is often used to describe a feeling of immense pleasure and relaxation, a sense of mental freedom and joy. It’s akin to feeling carefree and content, with one’s mind being refreshed and uplifted. This expression is frequently used to describe the pleasant emotions one experiences when in a beautiful, serene environment or when one is free from stress and worries.”

  1. Sanskrit Idiom

“The Sanskrit phrase “जलबिन्दुनिपातेन क्रमशः पूर्यते घटः” (jalabindunipātena kramaśaḥ pūryate ghaṭaḥ) means “a pot gets filled gradually by the falling of water drops.” It is a proverb that signifies the importance of patience and persistent effort. The idea is that just as a pot fills drop by drop, any large task can be completed by taking small, consistent steps. It encourages the belief that perseverance and gradual progression lead to the achievement of larger goals.”

Both Claude 2.1 and ChatGPT explain foreign language idioms in every way, but Claude 2.1 takes it a bit further by breaking down the idioms and explaining them with the translation.

4. Data Analysis

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👍👍👍 | 👍👍👍 |

ChatGPT and Claude 2.1 boast their abilities to break down complex documents. Claude, specifically with the recent updates, can scan technical documents and financial statements.

To test the data analysis capabilities, I fed Claude 2.1 and ChatGPT the 2023 Q4 financial statements released by Apple Inc.

Let’s see how they do,

Claude 2.1’s data analysis:

Claude 2.1 Data Analysis

I started a timer, and Claude 2.1 was much faster in analyzing the PDF document and providing a result. It broke down the key highlights of the data and also provided a takeaway for the document.

ChatGPT’s data analysis:

ChatGPT Data Analysis

ChatGPT took more time than Claude 2.1 and did not analyze the document. It was only upon further prompting that it broke down the specifics, for which it again took a substantial amount of time.

But, after further prompting and a little waiting, I got the results I needed.

5. Creativity

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👍👍👍👍 | 👍👍👍 |

As friendly as both chatbots are in the way they converse, they are equally creative.

ChatGPT has pretty good creativity compared to Claude 2.1. It can generate poems and stories with great creativity and deep meaning. While Claude 2.1 is also successful in generating poems and stories, they are less creative and lack deeper meanings.

Here are the responses generated by Claude and ChatGPT:

Poem on Starry Nights and AI by Claude 2.1:

Poem on Starry Nights and AI by Claude 2.1

**Poem on Starry Nights and AI by ChatGPT:

Poem on Starry Nights and AI by ChatGPT

**6. Coding Abilities

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👍👍👍👍 | 👍👍👍👍 |

To test the complexity of generating a code, I asked the chatbots to:

“Generate a code to create a basic website used for dropshipping.”

Here is the code generated by Claude 2.1 and ChatGPT:

By Claude 2.1:

Show Source Code </>

<html>
<head>
<title>My Dropshipping Website</title>
</head>
<body>
<header>
<h1>My Company Name</h1>
<nav>

<a href=”#”>Home</a>

<a href=”#”>Shop</a>

<a href=”#”>About</a>

<a href=”#”>Contact</a>

</nav>
</header>
<main>
<h2>Featured Products</h2> <section class=”products”> <div class=”product”>
<img src=”product1.jpg”>
<h3>Product 1</h3>
<p>$29.99</p>

<button>Add to Cart</button>

</div>
<div class=”product”>
<img src=”product2.jpg”>
<h3>Product 2</h3>
<p>$19.99</p>

<button>Add to Cart</button>

</div>
</section>
</main>
<footer>
<p>&copy;2023 My Company</p>
</footer>
</body>
</html>

By ChatGPT:

Show Source Code </>

<html lang=”en”>
<head>
<meta charset=”UTF-8″>
<meta name=”viewport” content=”width=device-width, initial-scale=1.0″>
<title>Simple Dropshipping Store</title>
<style>

body { font-family: Arial, sans-serif; }

.product { border: 1px solid #ddd; padding: 10px; margin: 10px; }

.product img { max-width: 100%; }

.product-title { font-size: 1.2em; font-weight: bold; }

.product-price { color: #E44D26; }

.add-to-cart { background-color: #4CAF50; color: white; padding: 10px 20px; border: none; border-radius: 5px; cursor: pointer; }

</style>
</head>
<body>
<h1>Welcome to Our Dropshipping Store</h1>
<div id=”product-list”>

<!– Products will be loaded here by JavaScript –>

</div>
<script>

// Assuming you have an API endpoint to fetch product data

const apiEndpoint = ‘https://example.com/api/products’;

// Function to fetch product data and display it

function loadProducts() {

fetch(apiEndpoint)

.then(response => response.json())

.then(products => {

const productList = document.getElementById(‘product-list’);

productList.innerHTML = ”; // Clear the product list

products.forEach(product => {

const productDiv = document.createElement(‘div’);

productDiv.className = ‘product’;

productDiv.innerHTML = `

<img src=”${product.image}” alt=”${product.title}”>
<div class=”product-title”>${product.title}</div>
<div class=”product-price”>$${product.price}</div>

<button class=”add-to-cart” onclick=”addToCart(${product.id})”>Add to Cart</button>

`;

productList.appendChild(productDiv);

});

});

}

// Function to handle adding a product to the cart

function addToCart(productId) {

console.log(‘Added to cart:’, productId);

// Here you would add logic to handle the cart process

}

// Load products on page load

loadProducts();

</script>
</body>
</html>

Both generated a working .html code to create a dropshipping website in a snap. Claude 2.1, however, went a little ahead with a better structure for the website, with examples of products and pricing added.

**7. Logical Reasoning

| ChatGPT | Claude 2.1 | | ----------- | -------------- | | 👍👍👍👍👍 | 👍👍👍👍👍 |

Next, I ran a classic logical reasoning from ‘Syllogisms’ to test their ability to break down complex reasoning questions.

I asked both chatbots the following question:

‘Two Statements are given, followed by four conclusions, I, II, III, and IV. You have to consider those Statements to be true, even if they seem to be at variance from commonly known facts. You are to decide which of the given conclusions can definitely be drawn from the given Statements. Indicate your answer.

Statements:

Some books are mobiles.

Some calculators are mobiles.

Conclusion:

I. Some mobiles are calculators.

II. Some mobiles are books.

(A) Only conclusion I follow

(B) Only conclusion II follows

(C) Both conclusions I and II follow

(D) Neither conclusion I nor II follows

The answer here is C.

Here’s what Claude and ChatGPT generated as answers.

Claude 2.1’s response:

Claude 2.1 Logical Reasoning

ChatGPT’s response:

ChatGPT Logical Reasoning

I did not feel a big difference comparing both for this parameter as both Claude 2.1 and ChatGPT perfectly understood the question and were also able to give out the correct solution for this logical reasoning test.

ChatGPT vs Claude 2.1: Capabilities

Let us look at the capabilities of ChatGPT and Claude.

ChatGPT Capabilities

Here are the capabilities of the ChatGPT model.

  • ChatGPT Plus provides access to API support and ChatGPT Plugins.
  • Supports more than 80 languages.
  • Can understand text inputs, analyze images and files, describe what is depicted in the images, and even generate an AI image.
  • Can answer queries related to any field of study. It has mastery in all the fields.
  • Cracks dad jokes which are funny enough to make the user smile.
  • Cracked most of the standardized examinations.
  • Does not respond easily to restricted texts.

Claude 2.1: Capabilities

Here are the capabilities of Claude AI Chatbot.

  • Claud earned a marginal pass on the blind-graded test.
  • Claude 2.1 has a token limit of 200,000 tokens.
  • It can have open-ended conversations, tell jokes, and discuss various subjects.
  • It can write fictional stories, poems, and other creative writings.
  • Sticks to its ethical principles, which makes it useful in industries like education, healthcare, and finance, where it needs to operate ethically.
  • Declines inappropriate requests frequently.
  • The writing style of the chatbot is more detailed and accurate.

ChatGPT vs Claude 2.1: Limitations

Every coin has two slides; similarly, there are limitations and capabilities in all the AI models.

Let us dive into the limitations of ChatGPT and Claude AI, including the areas where they have disappointed creators and users.

ChatGPT Limitations

The limitations of ChatGPT are as follows.

  1. ChatGPT still hallucinates factual information.
  2. Does not have access to web browsing and cannot provide real-time data. (GPT 3.5)
  3. ChatGPT needs additional prompting to get the result you want.

Claude 2.1: Limitations

Here are the limitations of Claude’s AI chatbot:

  1. Claude is weaker in complex mathematics when compared to ChatGPT.
  2. Does not have access to the internet.
  3. Claude also has the significance of generating inconsistent and false statements.

Which is better — Claude or ChatGPT?

ChatGPT and Claude 2.1 have extraordinary abilities and can help humans make some tasks easier.

While it was a fair tie when using both of them for their practical usage, it should be noted that Claude 2.1 is still in beta and cannot access the Internet yet.

It was unexpected that Claude performed better at coding than ChatGPT with fewer instructions, had more token sizes, and generated faster responses.

Sure, ChatGPT has had significant advancements and is based on a high-quality language model, but the constitutional AI from Antropic gives ChatGPT the run for its money with fewer features.

Seeing where Claude 2.1 goes with further updates will be interesting. As the new updates are rolled out, I will keep my content updated. So you can check out our content again!

FAQs

**Is ChatGPT the best AI chatbot?

ChatGPT is considered the best Chatbot yet, as it can flawlessly carry out the most difficult computational work.Moreover, it can understand images, and it can also process them. Besides, it can perform numerous tasks that other AI chatbots cannot.

**Who is a competitor of ChatGPT?

Google Bard AI is one of the most popular competitors of ChatGPT. It converses with the users in a Human-like tone and can generate responses like humans. Google recently released it, and only the selected users can access it.

**Which AI face generator is best?

Google Bard AI is one of ChatGPT’s most popular competitors. It converses with the users in a Human-like tone and can generate responses like humans.

**Is ChatGPT the most advanced AI?

ChatGPT is one of the most powerful and advanced AI. Moreover, with the recent release of GPT-4 and ChatGPT plugins, it has evolved and become unbeatable as it can generate code, understand and process visual inputs, summarize texts, etc.

요약하다
Anthropic's new AI competitor, Claude 2.1, is compared to ChatGPT in terms of features, capabilities, and performance. ChatGPT has seen significant improvements with new features like plugins and code interpreter, while Claude 2.1 boasts a larger token limit and improved comprehension. ChatGPT offers better creativity, code generation, and humor, while Claude 2.1 excels in factual understanding and logical reasoning. ChatGPT is more widely available and supports more languages compared to Claude 2.1. Both models use reinforcement learning techniques, but ChatGPT seems to have an edge with its recent updates and features. The token size limit for ChatGPT ranges from 4,096 to 8,192 tokens, while Claude 2.1 can handle up to 200,000 tokens. Processing time for ChatGPT is around 30 seconds, while Claude 2.1 takes about 5 seconds. There are differences in their knowledge bases, with ChatGPT having a knowledge cutoff in January 2022 and Claude 2.1 claiming to include the latest information. Overall, ChatGPT appears to offer a more comprehensive and versatile experience compared to Claude 2.1.