Claude 2.1: 200K context and fewer hallucinations
In one sentence Anthropic ships Claude 2.1: 200K-token context window (~500 pages), 2× reduction in false statements on borderline questions, tool use in beta. Reply to GPT-4 Turbo 128K.
Two weeks after GPT-4 Turbo (128K), Anthropic raises the bar: Claude 2.1 brings the context window to 200K tokens, equivalent to about 500 pages of text. That means you can load an entire legal contract, a technical book, or a large codebase into a single request and ask questions over it.
The other key improvement is fewer hallucinations. Anthropic states Claude 2.1 produces 50% fewer "confident false answers" on questions where the right answer is "I don't know". It's the first step toward what will become a Claude differentiator: honestly saying it doesn't know instead of making things up.
Tool use also arrives in beta (OpenAI-style function calling) along with dedicated system prompts. Pricing: $8/M input, $24/M output — more expensive than GPT-4 Turbo. Positioning is clear: Claude for enterprise cases where long context and reliability matter more than price.
Companies
Anthropic
Tools
Claude 2.1
Tags
Sources