Claude 2.1: Anthropic's AI Update Promises Speed, Smarts and Savings

Anthropic has upgraded its Claude large language mode, boosting its ability to read lengthier documents.

The Amazon and Google-backed startup unveiled Claude 2.1 which boasts a 200,000 token context window – roughly around 133,000 words or 533 pages. Claude 2.1’s improved context window now puts it ahead of GPT-4 Turbo, OpenAI’s newest model that can handle 128,000 tokens.

Users can now upload technical documentation like entire codebases, financial statements or lengthy literary works that Claude can analyze in mere minutes.

However, the improved 200,000 token context window is reserved solely for Claude Pro users, which comes with a $20-a-month subscription.

Among the other updates to Claude include a huge reduction in the rate of hallucinations – Anthropic says Claude 2.1 boasts a two-times reduction in false statements compared with Claude 2.0.

The company claims the reduction in hallucinations means businesses can use Claude 2.1 to “build high-performing AI applications that solve concrete business problems and deploy AI across their operations with greater trust and reliability.”

Claude 2.1 also boasts improvements in comprehension and summarization, particularly for long, complex documents. The team behind it claims it offers a 30% reduction in incorrect answers and a three to four times lower rate of mistakenly concluding a document supports a particular claim.

Related:OpenAI Sought Merger with Anthropic; Sam Altman the ‘Martyr’

Claude can now be integrated with users’ APIs – meaning Claude can be incorporated into existing applications and workflows.

The model can also search over web sources and retrieve information from private knowledge bases – a move to rival ChatGPT’s ability to access real-time information.

Anthropic has also introduced system prompts, which allow users to provide custom instructions to Claude to improve performance.

How to access Claude 2.1

Claude 2.1 is available in Anthropic’s API. It’s also powering the chat interface at claude.ai for both the free (bar the 200k context length) and Pro tiers.

Pricing updates

The unveiling of Claude 2.1 comes with a pricing restructure – offering generations cheaper than OpenAI’s top model.

Claude 2.0 and 2.1 cost $0.008/1K token inputs and $0.024/1K token outputs. By comparison, OpenAI’s GPT-4 Turbo costs $0.01/1K token inputs and $0.03/1K token outputs.