Why did Phi give an incorrect or unexpected response?

Phi is built on top of AI models created by OpenAI, Anthropic, Google, and other AI providers. While these models are powerful, they can sometimes "hallucinate" or provide inaccurate information. This is a known limitation of current AI technology.

While we use various strategies to improve the response within Phi, no AI can guarantee perfect responses. Here are a few tips:

• Turn on Thinking: This helps the AI reason through its logic more carefully. Use the "Deep" setting for complex queries.

• Be specific about what you need, and what types of sources you would like Phi to reference.

• Verify important information: double-check anything critical.

• If you're using a model that costs less Energy, try to select one of the top models instead.

Want to help us improve? If you'd like to report a specific response, tap the 👍 / 👎 options below any message. You can attach the message exchange and send it to our team for review.