Edge AI News Intelligence

Ask about the news

Fast Haiku 5.5 Capable Opus 5.5

Under the hood

Web search When on, Claude can search the web (via You.com) for information the stored articles don't cover, up to 3 searches per answer. Answers get slower and use more tokens.

Per-response details

Ask a question to see what each answer used.

#	Model	DB articles	Article context (≈ tokens)	Web searches	Input tokens	Output tokens	Total

Input and output tokens are exact counts reported by the API, summed over every model call in an answer (a web search adds a call). Article context is an estimate of how much of the input the database articles took up.

Answers come only from the articles stored here, with numbered citations.

How relevance scores work

Claude scores each article 1–10 for how useful it is to someone building on-device AI products, using this rubric. Hover a score, or read the “Why” line on a card, for the reason behind it.

New marks articles added by the most recent refresh.