![Anthroic]()
Anthropic has introduced Citations, a new feature for its developer API designed to enhance the reliability and accuracy of AI-generated responses. Citations allow developers to “ground” answers from Anthropic’s Claude AI models in specific source documents, such as emails, by referencing the exact sentences or passages used in generating the response.
As of January 2025, Citations is integrated into Anthropic’s API and Google’s Vertex AI platform. This feature is especially valuable for tasks such as document summarization, Q&A systems, and customer support, where detailed and transparent references are critical. Developers can now ensure their AI outputs are backed by credible, specific sources, improving trust in their applications.
![Official post]()
Supported Models and Cost Considerations
Initially, Citations were supported only by the Claude 3.5 Sonnet and Claude 3.5 Haiku models. The feature is not free and incurs charges based on the length and number of source documents processed. For instance:
- A ~100-page document costs approximately $0.30 with Claude 3.5 Sonnet.
- The same document costs around $0.08 with Claude 3.5 Haiku.
![Models]()
This pricing reflects Anthropic’s standard API rates and offers an affordable solution for developers who want to minimize AI hallucinations and errors.
Driving Transparency and Accountability
By enabling models to cite their sources, Citations help reduce the risk of unsupported or fabricated claims in AI responses. Anthropic envisions this feature as a critical step toward improving the trustworthiness and accountability of AI-generated outputs.
A Tool for Developers and Businesses
Citations are expected to be a game-changer for developers and businesses looking to integrate AI into workflows where accuracy and transparency are paramount. Whether summarizing documents, answering customer inquiries, or supporting decision-making processes, the feature offers a clear advantage by rooting AI-generated insights in verifiable data.