Less than 6 months after debuting its ChatGPT rival, Claude, Anthropic is rolling out an updated version. And this time, the company is promising the chatbot will provide more detailed answers, fewer harmful responses, improved reasoning behind answers, and simply overall better performance than before.
Also: How to use ChatGPT: Everything you need to know
The company says Claude 2 is easier to converse with and better explains its thinking -acting more like a real human.
Available in the United States and the UK today, Claude 2 packs a lot more power than its predecessor.
In the world of AI, user input is broken down into "tokens," or bits of information broken down for the AI to process. The updated Claude can process 100,000 tokens, a significant upgrade from the 9,000 of the previous version.
That means Claude 2 can understand and process over 70,000 words at once, leading to better and longer responses. The AI can process hundreds of pages of documents, Anthropic says, and output much longer documents, including short stories.
Also: Six skills you need to become an AI prompt engineer
Further showing how intelligent its new AI is, Anthropic noted the latest model of Claude scored a 76.5% on the multiple-choice portion of the bar exam, an improvement from the previous version's score of 73%. New Claude also brings greatly improved coding skills, Anthropic claims. Its original version scored a 56% on the Codex HumanEval (a Python coding test) while the new version jumped to a 71%.
Claude 2 is also significantly safer. Anthropic admits that "no model is immune from jailbreaks," but says it's harder to prompt its new model to issue an offensive or dangerous response. In fact, the company said, according to its manual analysis, Claude 2 was twice as good at giving safe responses.
The new AI can be accessed via API or through a dedicated site, claude.ai.