Anthropic, the AI business co-founded by former OpenAI workers, has just launched Claude Pro, its first consumer-facing premium subscription plan for Claude 2, the company’s AI-powered chatbot that analyses text. Customers receive “5x more usage” than the free Claude 2 tier, the capacity to send “many more” messages, priority access to Claude 2 during times of high traffic, and early access to new features for the monthly fee of $20 in the U.S. or £18 in the U.K.
The cost of Claude Pro is the same as that of OpenAI’s ChatGPT Plus, a paid subscription to ChatGPT, a Claude 2 competitor. “Since launching in July, users tell us they’ve chosen Claude as their day-to-day AI assistant for its longer context windows, faster outputs, complex reasoning capabilities, and more,” Anthropic noted in a blog post. “Many also mentioned how much they would enjoy lengthier talks and greater file submissions…
Subscribers now have 5 times more usage of our newest model with Claude Pro. Every eight hours, when the message limit refreshes, Anthropic estimates that Claude Pro users will send at least 100 messages to Claude 2. (This contrasts with the ChatGPT Plus membership limit of 50 messages per three hours.) Why the cap?
Constrained capacity, according to Anthropic, is as follows: When responding to massive attachments and lengthy talks, Claude 2 requires a lot of strong machines to run. We imposed these restrictions so that Claude could be made widely available for free use while enabling power users to include Claude in their regular workdays.
To address Anthropic’s concern, hosting AI chatbots like Claude 2 is expensive. According to reports, OpenAI once paid $700,000 per day, or around $21 million per month, to run ChatGPT. Longer chats, particularly those with heavy attachments, cause the Claude 2 message limit to be consumed more quickly. For instance, after uploading a copy of “The Great Gatsby,” a Claude Pro member would only be able to send about 20 more messages during the following eight hours. That’s because each time Claude 2 receives a message, it “re-reads” the entire conversation, including any attachments.
Anthropic’s ultimate goal, as we’ve previously covered, is to develop a “next-gen algorithm for AI self-teaching,” as it puts it in a recent presentation deck to investors. A virtual assistant that can answer emails, conduct research, create art, books, and other tasks could be created with such an algorithm. We have already seen some of these capabilities in huge language models like GPT-4.
Anthropic, which was founded in 2021 and is run by Dario Amodei, a former vice president of research at OpenAI, has so far received $1.45 billion at a valuation of a few billion. While that may seem like a lot, the business projects that it will require $5 billion over the next two years to build the AI that it has in mind. The majority of the funds, including sales of items like Claude Pro, will be used to expand the available computing power.
Anthropic suggests in the deck that it uses “tens of thousands of GPUs” in clusters to train its models and that it will need to invest about a billion dollars on infrastructure in the upcoming 18 months alone. Quora, which offers access to Claude 2 and Claude Instant, a less functional but less expensive version of Claude 2, through its subscription-based generative AI app Poe, is one of Anthropic’s alleged “thousands” of current clients and partners. Cohere, AI21 Labs, and OpenAI, whose AI tools are utilized by millions of developers and which supposedly plan to earn $1 billion in revenue next year, are all fierce competitors.
