Anthropic has released Claude 2.1, a large language model with a 200,000-token context window, outperforming OpenAI’s 120K context. This strategic release, a result of a partnership with Google, offers context-handling prowess nearly double that of its closest rival, responding to the growing demand for AI for long-form document analysis.
Did you come here looking for “ChatGPT Signup Is Currently Unavailable”? Whe also have a post about that: How to Fix “ChatGPT Signup Is Currently Unavailable” & Alternatives When It’s Down
Claude 2.1, an AI model, has processed a 200K message, an industry first. The technology is expected to reduce latency and increase honesty, with a 2x reduction in false positives compared to the previous Claude 2.0 model.
As a result, organisations can build high-performance AI applications that solve business problems and deploy AI with greater confidence and reliability. Claude 2.1’s honesty was tested by curating complex, factual questions to explore known weaknesses in current models.
Using a rubric that distinguishes false claims from admissions of uncertainty, Claude 2.1 was found to be more likely to demur than provide false information. This advancement enables organisations to build powerful AI applications that solve business problems, and deploy AI across their operations with greater confidence and reliability.
Anthropic’s Claude 2.1 has improved accuracy by 50% and doubled truthfulness compared to Claude 2.0. The improvements were tested against complex, factual questions to challenge the limitations of current models. Hallucinations were previously a weakness of Claude, and the dramatic increase in accuracy would put the LLM in closer competition with GPT-4.
Claude 2.1 also integrates more seamlessly into advanced users’ workflows with an API tool usage feature, demonstrating its ability to orchestrate across functions, search the web and pull from private databases. This feature promises to extend Claude’s utility across multiple operations.