Prompt with Claude
Understanding Claude as an AI Assistant
Last updated
Was this helpful?
Understanding Claude as an AI Assistant
Last updated
Was this helpful?
We are big fans of Claude at AirOps. However, here are key differences to consider with Claude:
Claude is generally faster than OpenAI models.
Claude has a larger context window available (200K tokens).
Claude 3 has demonstrated superior performance in benchmark cognitive tests over previous models and even surpasses OpenAI's GPT-4 in many areas.
Claude is chatty - it tends to include a "Chat" response such as "Sure! Here is a ...".
According to , Claude best understands XML tags.
In order to prompt with Claude, consider adding <task></task>.
Claude 3 offers a family of models (Haiku, Sonnet, Opus) with varying capability levels. Choose the model that best suits your needs based on the trade-off between:
Capability: Higher-tier models (Sonnet, Opus) handle complex tasks better.
Speed: Haiku, the base model, is the fastest.
Cost: Higher-tier models may incur higher usage costs.
To convert our , we add XML tags around the task and the URL:
System: You are an expert at url information extraction.
User: <task>Extract the unique transcript ID from the end of the URL.</task><url>https://app.fireflies.ai/view/AirOps-Intro::DkMo5PetdLeCDglp</url>
Assistant: DkMo5PetdLeCDglp
User: <task>Extract the unique transcript ID from the end of the URL.</task><url>https://app.fireflies.ai/view/Mdei9OrqpQlWIijd</url>
Assistant: Mdei9OrqpQlWIijd
User: <task>Extract the unique transcript ID from the end of the URL.</task><url>{{link_to_sales_call}}</url>
For more information on how to prompt with Claude, check out: