Claude 2.1 introduces 200K Context Window

Just days after the turmoil at OpenAI, which has cast doubt over the future of ChatGPT, Anthropic has released an update to its own AI chatbot, Claude 2.1. This latest upgrade to Anthropic’s flagship model not only challenges OpenAI’s latest language models but also introduces unique features that set it apart.

What is Claude?

Claude, launched in March 2023 by Anthropic (established in 2021 by Dario and Daniela Amodei), works similarly to ChatGPT. Claude can help with use cases including summarization, search, creative and collaborative writing, Q&A, coding, and more.

What sets Claude apart is its focus on safety. Anthropic has made a conscious effort to design Claude in a way that minimizes the risk of harmful outputs, such as bias, misinformation, and offensive language.

Claude powers two popular web apps: Quora’s Poe and Notion.

What’s new in Claude 2.1?

The update includes three major changes:

1. Larger context window
2. Reduced hallucination rates
3. “API tool use” for developers

Larger context window

Claude 2.1 has doubled the context window to 200K tokens, which is significantly larger than the 128K token context window of OpenAI’s GPT-4. The 200K context window translates to roughly 150,000 words, or over 500 pages of documents.

It means you can upload an entire book and talk to it.

OpenAI’s GPT-4 currently has 128K context windows, so that means Claude can handle 62K more tokens.

Note: The full 200K token capacity remains exclusive to paying Claude Pro subscribers for now. Free users will continue to be limited to Claude 2.0’s 100K tokens.

Reduced hallucination rates

Anthropic claims that Claude 2.1 has reduced the AI’s hallucinations and false statements by up to 50%. This is a significant improvement, as hallucinations can be a major problem with large language models.

In an internal test performed by Anthropic, Claude 2.1 was significantly more likely to demur rather than provide incorrect information compared to Claude 2.0.

In terms of comprehension and summarization, Claude 2.1 demonstrated a 30% reduction in incorrect answers and a 3–4x lower rate of mistakenly concluding a document supports a particular claim.

“API Tool Use” for developers

Users can now connect API tools, and Claude will pick the best one for the job depending on the context. It means the AI will decide which tool is required to achieve the task and execute an action on their behalf.

This works similarly to the automatic tool switching feature from ChatGPT.

How to access Claude

Claude is available for free at the Claude website. The company recently expanded its support to 95 countries.

A pro-tier subscription is also available for $20 per month, which provides several benefits, including:

• 5x more usage than the free tier
• Priority access during high-traffic periods
• Early access to new features (That includes the new 200K Context Window)

Final Thoughts

While the new features are not groundbreaking, they are nonetheless welcome improvements. It’s encouraging to see companies like Anthropic continue to push the boundaries of AI technology, especially at a time when large players like OpenAI are facing internal challenges.

Claude’s focus on safety and reliability remains a key differentiator, particularly in light of the recent concerns surrounding OpenAI.

If you’re a developer looking for an alternative to OpenAI’s API, I recommend using Claude.

‍