Anthropic's ChatGPT Alternative Claude 2 Has an Awesome New Feature and is Now Available to Everyone

Anthropic also announces Jasper AI as a customer

Anthropic has introduced Claude 2 in open beta. Claude 2 is a large language model (LLM) powered ChatGPT alternative. The predecessor Claude 1 was only available in closed beta and through a few partner organizations such as Quora’s Poe or as a restricted-access API. So, Claude 2 will be the first exposure for most people to Anthropic’s chatbot and LLM.

Claude 2 offers a familiar chat interface that, through color and other minimalistic design choices, carries a different appearance than ChatGPT. Anthropic also says it has better performance than the most recent 1.3 model version. According to the announcement:

We have made improvements from our previous models on coding, math, and reasoning. For example, our latest model scored 76.5% on the multiple choice section of the Bar exam, up from 73.0% with Claude 1.3. When compared to college students applying to graduate school, Claude 2 scores above the 90th percentile on the GRE reading and writing exams, and similarly to the median applicant on quantitative reasoning.

Think of Claude as a friendly, enthusiastic colleague or personal assistant who can be instructed in natural language to help you with many tasks. The Claude 2 API for businesses is being offered for the same price as Claude 1.3. Additionally, anyone in the US and UK can start using our beta chat experience today.

…

In addition, our latest model has greatly improved coding skills. Claude 2 scored a 71.2% up from 56.0% on the Codex HumanEval, a Python coding test. On GSM8k, a large set of grade-school math problems, Claude 2 scored 88.0% up from 85.2%.

100K Token Context Window

Anthropic is currently the king of the context window. It enables users to upload as many as 100k data tokens which Anthropic says is about 75,000 words. This is the context that Claude 2 can maintain. It means you can upload book-length text content and ask the chatbot about the contents. It will answer based on the context (i.e., text input) you provide as well as the capabilities it has from the LLM training.

Users can input up to 100K tokens in each prompt, which means that Claude can work over hundreds of pages of technical documentation or even a book. Claude can now also write longer documents - from memos to letters to stories up to a few thousand tokens - all in one go.

The large context window also is beneficial if you have a lengthy interaction with Claude. It will remember the earlier parts of your chat and keep that context in subsequent responses.

OpenAI’s GPT-4 has a 32k context window, but it is only available through the API and is restricted access. The standard ChatGPT context window runs off the 4k context window driven by the GPT-3.5 model. GPT-4 has a standard context window of 8k. Claude 2’s context window is nearly eight times as large as the standard GPT-4 deployment and twice that of the standard model.

Document Upload

Another notable difference from ChatGPT is the ability to upload documents. This is useful for summarization, but also for asking questions and transforming the content.

From a summarization standpoint, Claude does what you would expect. It can provide bullet points or text summaries of varying lengths. It can also answer specific questions about the contents of the report or create a quiz to test your recall and reading comprehension.

Everyone is enamored with the ability of LLMs to answer questions about a wide variety of topics. Summarization of text is not as widely used today but is certainly one of the most valuable features offered by LLMs. While you need to go through an elaborate cut-and-paste method to achieve these results with ChatGPT, Anthropic is making the process simple by supporting document uploads.

No Real-Time Internet Access

Max Verstappen won the Monaco Grand Prix 2023. It has occurred. Claude 2 doesn’t know this because it was most recently trained before May 2023. It also doesn’t know about its existence since it launched in July or that Google Bard is now running on the PaLM AI model, not LaMDA.

However, Claude did correctly identify the launch of Bing Chat in February, suggesting the most recent training for Claude on web data was likely in March or April 2023. This will not be your replacement for search. It will take on similar tasks that you may employ ChatGPT or Google Bard for today.

Modeling Creativity

Claude 2 also deftly navigates creative writing. When asked to write about the history of LLMs in the style of Dr. Seuss, it delivered a response that you might expect from ChatGPT’s GPT-4 model.

Jasper Adds Claude

The addition of Jasper AI among thousands of API customers is another noteworthy element of the new announcement. Jasper AI is an AI writing assistant solution that was built originally on OpenAI’s GPT-3.

When I spoke with company president Shane Orlick on the Vociebot Podcast in January, he said that Jasper was looking to offer additional models beyond OpenAI’s offerings. According to Anthropics announcement:

We are also currently working with thousands of businesses who are using the Claude API. One of our partners is Jasper, a generative AI platform that enables individuals and teams to scale their content strategies. They found that Claude 2 was able to go head to head with other state of the art models for a wide variety of use cases, but has particular strength for long form low latency uses. "We are really happy to be among the first to offer Claude 2 to our customers, bringing enhanced semantics, up-to-date knowledge training, improved reasoning for complex prompts, and the ability to effortlessly remix existing content with a 3X larger context window," said Greg Larson, VP of Engineering at Jasper. "We are proud to help our customers stay ahead of the curve through partnerships like this one with Anthropic."

Hot Take

Claude 2 is a very capable LLM-based chatbot in the same league as ChatGPT and Google Bard. It is going to be hard for Anthropic to generate the same type of enthusiasm as ChatGPT because there are many options today. Claude 2 shows a similar solution in most regards.

However, the key for LLMs today is to differentiate based on domains, features, or cost. This is true whether the LLM is driving demand for a consumer chat application or API access. OpenAI has the most awareness, mindshare, and trust in the market for general-purpose LLMs. What will compel users to try another solution? A unique or better-performing feature is critical.

Earlier this year and in 2022, Anthropic promoted its constitutional AI training approach that it says delivers “safer” and “better aligned” model outputs. There is no evidence that this is true, but it may be true and is a worthwhile goal.

The more interesting angle is that few people outside of Anthropic employees talk about this concept. And Anthropic has barely mentioned it lately. I suspect that is because constitutional AI didn’t generate enough interest among potential users. It may be a differentiated product attribute, but not a catalyst for adoption to date.

Anthropic’s differentiation today is about summarization and document information inquiry. The large context window combined with the ability to upload documents will definitely drive adoption among the segment of users that needs this feature. It is the hook to induce trial that could lead to broader use of general-purpose features as well.

What do you think? Do you like Claude? Will you start using it instead of ChatGPT or Bard? Let me know in the comments.