Nov 22, 2023 10:53:00

Anthropic releases large-scale language model 'Claude 2.1', can read up to 200,000 tokens and 150,000 words, reduces hallucinations by half, provides new API integration, etc.

Anthropic, an AI startup founded by former OpenAI engineers, has announced Claude 2.1, a large-scale language model. Claude 2.1 has doubled the number of tokens it can process, now accepting up to 200,000 tokens, and has significantly improved accuracy.

Anthropic\Introducing Claude 2.1

https://www.anthropic.com/index/claude-2-1

Our new model Claude 2.1 offers an industry-leading 200K token context window, a 2x decrease in hallucination rates, system prompts, tool use, and updated pricing.

Claude 2.1 is available over API in our Console, and is powering our https://t.co/uLbS2JNczH chat experience. pic.twitter.com/T1XdQreluH

— Anthropic (@AnthropicAI) November 21, 2023

Claude 2.1 can process 200,000 tokens, which is significantly more than the approximately 32,000 tokens that OpenAI's 'GPT-4' can process. This equates to approximately 150,000 English words and over 500 pages of material. Being able to process large amounts of content and data at once expands the range of things you can do. Anthropic describes its large-scale language model that can process 200,000 tokens as an 'industry first.'

In addition, the occurrence of so-called ' hallucinations ', the phenomenon of 'lying' seen in large-scale language models, has been reduced by half compared to 'Claude 2.0'. When Anthropic tested Claude 2.1 to find weaknesses, Claude 2.1 was more likely to give up than provide inaccurate information. Anthropic says it will continue to work on improving the accuracy and reliability of its output.

With the release of Claude 2.1, Anthropic also started offering tools as new features in the beta version. Users will be able to search databases, use web search APIs to answer questions, and perform actions in the software via private APIs. At the time of announcement, the use of the tool is in the early stages of development, and functions and guidelines for developers are currently being created.

Claude 2.1 is available via API at the time of article creation, and can be accessed from both the free and professional versions of the chat service

claude.ai . However, processing up to 200,000 tokens is limited to Claude Pro users. Anthropic says, 'We are building the most secure and technologically sophisticated AI system in the industry. We can't wait to see how our new features will inspire and use cases. ” he said.