Anthropic, the San Francisco-based AI startup, has announced an upgrade to its language model called Claude 2.1. The new model features a 200,000-token context window, which is nearly double the size of OpenAI’s recently released GPT-4 model.
The upgrade is a result of Anthropic’s expanded partnership with Google, which provides access to advanced processing hardware. Claude 2.1 has the ability to process lengthy documents and accurately grasp information from longer prompts compared to GPT-4.
The model also offers a 50% reduction in hallucination rates and new features such as an API tool for advanced workflow integration. However, the full 200K token capacity is currently only available to paying Claude Pro subscribers.