Claude 2 raises the stakes in the AI model segment.
Anthropic, an AI startup with Google’s backing, has launched its latest AI model called Claude 2. This upgraded version offers enhanced performance, longer responses and accessibility through an application programming interface (API), all at the same price as its predecessor, Claude 1.3. According to the company, Claude is designed to be a friendly and enthusiastic colleague or personal assistant who can understand natural language instructions and assist with various tasks. The upgraded version of Claude is now available to general users in the United States and the United Kingdom through the beta-test website claude.ai.
As the successor to Anthropic’s Claude 1.3, which debuted earlier this year, Claude 2 brings significant improvements in coding, math and reasoning capabilities. Notably, the new model achieved a score of 76.5% in the multiple-choice section of the Bar exam, surpassing Claude 1.3’s score of 73.0%. Furthermore, when compared to college students applying for graduate school, Claude 2 performs above the 90th percentile in the Graduate Record Examinations (GRE) reading and writing exams, while displaying similar aptitude to the median applicant in quantitative reasoning.
Generative AI platform Jasper and code AI platform Sourcegraph are among the popular names that have already started testing the capabilities of Claude 2, exploring how it can be integrated into their workflows and systems.
Enhanced capacity
In a single prompt, the new model can process up to 100,000 tokens (approximately 75,000 words). This capability allows Claude 2 to handle extensive content, including technical documentation spanning hundreds of pages or even an entire book. The chatbot can now generate longer documents seamlessly, ranging from memos and letters to stories, extending up to several thousand tokens, without the need for multiple prompts.
Claude 2 also showcased enhanced coding skills, achieving an impressive score of 71.2% on the Codex HumanEval Python coding test, indicating its effective understanding and writing of code. This represents a significant advancement compared to Claude 1.3, which scored only 56.0% on the same test. Claude 2 also achieved a commendable score of 88.0% on GSM8k, a comprehensive collection of math problems typically taught in grade school, surpassing Claude 1.3’s score of 85.2%.
Safety first
Safety has been a paramount focus for Anthropic in developing Claude 2. Extensive measures, including automated tests and manual checks, have been undertaken to enhance the model’s safety features. As a result, Claude 2 has demonstrated twice the effectiveness of Claude 1.3 in producing harmless responses, making it less prone to generating offensive or dangerous output.
With Claude 2’s emergence, the AI model landscape gains a promising contender. While ChatGPT has established itself as a leading AI language model, the impressive features and advancements of Claude 2 offer a compelling alternative. The market’s response to Claude 2’s capabilities and pricing will undoubtedly be intriguing as businesses and developers assess its potential and ability to effectively compete with the established presence of ChatGPT.
Also read:
- The Chatbot Revolution: Meet the Top Up-and-Coming Competitors to OpenAI’s ChatGPT
- Google Bard vs. ChatGPT: Which Is the Better AI?
- 5 Essential Reasons Chatbots Fail—and Will ChatGPT, Too?
Header image courtesy of Pexels
Press Release link: https://www.anthropic.com/index/claude-2
