Anthropic Launches Claude 2 AI Chatbot with Enhanced Coding and Reasoning
Anthropic releases Claude 2, an upgraded AI model that shows significant improvements in coding, math, and standardized testing. The chatbot is currently available in beta across the U.S. and U.K. through a web interface and a paid API.
Anthropic launches Claude 2, the second generation of its commercial AI chatbot, making it available in beta to users in the U.S. and U.K. The new model operates through a web interface and a paid API, keeping the same pricing structure as its predecessor at roughly $0.0465 per 1,000 generated words. Several companies, including Jasper and Sourcegraph, already pilot the new system to test its capabilities in real-world applications.
The upgraded model outperforms Claude 1.3 across various academic and technical benchmarks. Claude 2 achieves a 76.5% score on the bar exam multiple choice section, passes the U.S. Medical Licensing Exam, and scores 71.2% on a Python coding test compared to the previous 56%. Anthropic attributes these enhancements to improved reasoning, better instruction-following skills, and training on more recent data from early 2023 that includes non-English text.
Despite the performance gains, Claude 2 remains architecturally similar to the previous version rather than representing a completely new foundation. Anthropic describes the release as the result of a continuous, iterative development approach over two years. Unlike some competing models, Claude 2 still lacks the ability to search the live web, focusing instead on processing provided documents, summarizing, writing, and answering topic-specific questions.