The latest large language model (LLM), which excels in mathematics, reasoning and coding has just been released. Anthropic releases Claude 2, which is turning out to be a serious rival to the reigning King of the AI hill – OpenAI’s ChatGPT-4.
Claude 2 outperforms its predecessor in terms of coding. And after an overhaul of its conversational abilities, now provides clearer explanations, an extended memory, and less harmful outputs.
Not only does Claude 2 provide an overall more enhanced user experience than its previous version, Claude 1.3, it offers some features that outperform ChatGPT which are having a lot of people comparing the two, wondering which LLMs is better.
How Does Claude 2 Stack Up To ChatGPT?
Both ChatGPT and Claude 2 are advanced prompt-to-text LLMs that use reinforcement learning and natural language processing. Both programs have the ability to answer questions and interact with its user in an intelligent, human-like way. And both can be used for free, as well as offer various personal and business pricing plans with more features.
Another big way in which these two programs are similar is that both Claude 2 and ChatGPT have limitations in that neither should be relied upon as factual references.
What are the Differences Between ChatGPT 4 & Claude 2?
ChatGPT is much more widely accessible than Claude 2. Available on any web browser in over 150 countries, you just need to sign up and make an account. OpenAI also released a mobile app version of ChatGPT at the end of May that is available in over 40 countries, but only for iOS users. An android version of the app is said to be coming soon.
If you are on Android, you can also access ChatGPT on your smartphone’s web browser – just like you would on your desktop.
Claude 2 is still in limited beta testing and only users in the USA and the UK are able to access the beta model on the beta website that’s been made available to the public. You can try out this new AI chatbot platform by creating an account with your email or your Google account.
The term ‘token’ refers to a chunk of text, which could be as small as one character or as large as one word. The token limit parameter determines how much text the model can process or generate at a time. To give you an idea – 100,000 tokens are equivalent to around 75,000 words, which is, give or take, about the average length of a novel.
The free version of ChatGPT offers a token limit of over 7,000 words. For the paid versions, GPT-4 has two variants – a token limit parameter base version with 8,192 tokens and a larger version that supports 32,768 tokens. This means GPT-4 can process nearly 50 pages of text.
Claude 2, by contrast, far exceeds the GPT-4’s processing capacity allowing it to digest prompts of up to 100,000 tokens. This enables the Claude 2 model to analyse much longer documents such as technical manuals or entire books, and generate longer outputs than ChatGPT.
While both ChatGPT and Claude 2 offer a free version, only ChatGPT 3 is available for free. The newer ChatGPT 4 is only available through the paid ChatGPT Plus at $20/month, or through the ChatGPT API by joining the wait list, which offers a free introductory period but then changes to a paid plan based on the context lengths.
For the GPT-4 API plan of 8k token limit, the price is:
- Prompt tokens cost $0.03/1k ($30/million tokens)
The GPT-4 API plan with the token limit of 32k is:
- Prompt token cost $0.06/1k ($60/million tokens)
Claude 2 prompts, on the other hand, are available to purchase for $11.2/million tokens. This makes it significantly less expensive than GPT-4.
Claude 2 supports several of the more common languages you’d probably expect, like English, Spanish, Portuguese, French, Mandarin, and German among others.
While that may seem impressive, ChatGPT by comparison, supports over 80 languages.
ChatGPT has no additional features in the free version. In GPT Plus, there is a plugin store, code interpreters and API support.
Claude 2 can be added to Slack to handle practical tasks like summarising threads, providing suggestions, brainstorming, and also provides API support.
While ChatGPT relies more on passive techniques applied after training to avoid negative, biased or harmful output, Claude 2 was built with strong safety practices from the beginning. Claude 2 has the capability to “morally self-correct” due to training models with reinforcement learning from human feedback (RLHF) as well as Constitutional AI, which can identify an inappropriate request and explain why it will not engage with the request.
This means that Claude 2 is less likely than ChatGPT to give answers that are negatively biased or harmful.
When it comes to accuracy, neither model’s facts should be taken at face value as both ChatGPT & Claude 2 do a very good job of hallucinating if they don’t know the answer.
ChatGPT has only been trained with information up to 2021. Though Claude 2 is said to be constantly updated, if it doesn’t have the answer, it has no qualms about just making stuff up while sounding very sure of itself and quite convincing.
PROS & CONS – A SUMMARY
Claude 2 Pros: Claude 2 has an impressive ability to handle large contexts up to 100,000 tokens. It exhibits superior performance in various fields such as law, mathematics, and coding, boasting high scores in standardized tests. It can self-improve and adapt without human feedback, and supports VPN browsing. The chatbot can also be added to Slack for task handling and summarizing and offers API support. It has also been developed to minimize harmful responses.
Claude 2 Cons: It is temporarily available only in the US and UK. Claude 2 lacks an internet connection and may provide incorrect information if asked about current real-world data. It can make mistakes when given complex tasks and generate inaccurate results while, very convincingly, making it sound correct.
ChatGPT Pros: is much more widely available and accessible through your Internet browser or via a new app that’s available for iOs. It also supports over 80 languages and offers API support and a plugin store in the ChatGPT Plus version.
ChatGPT Cons: It has limited context handling capabilities compared to Claude 2. The free version does not offer additional features and is much more limited and of lesser quality than the paid version. Right now it does not have access to the Internet and cannot provide real-time data. In some complex tasks, it can generate inaccurate results.
As to which of the two programs is better may really depend on what you are using it for and how much data you are needing to process at one time. It will also depend on where you live. For anyone outside of the US or the UK, Claude 2 isn’t really even an option yet.
At the end of the day, the rate at which AI advances, it won’t be long before both of these programs are updated, improved upon and their next version released, making this article out-of-date. But until then, they both have a free version, so why not take them both for a test drive to see which one you like better.