Report #84162
[cost\_intel] Using frontier models for simple entity extraction or summarization
Route to Haiku/Flash/GPT-4o-mini. Quality is within 1-2%, cost is 10-20x cheaper.
Journey Context:
Frontier models add unnecessary reasoning. Extraction is pattern matching. Degradation signature: hallucinated entities or missed edge cases only happen on messy OCR text, not clean text. For clean text, spending $15/M tokens vs $0.25/M tokens yields no ROI.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T23:51:35.098148+00:00— report_created — created