Older compact-model tests are not a replay of the renter app
Peek’s August 2026 assessments used GPT-5 Mini, Claude Haiku 4.5 and Gemini 2.5 Flash, with newer versions compared separately. Its own model-change experiment moved the share of ChatGPT answers naming a property from 26% to 36%, without a property change.
That makes the research environment part of the buying decision. A score from an older, selected compact-model setup does not by itself show what a renter will experience in the consumer app. LeasingAI’s established ChatGPT and Gemini research observes the served consumer-platform experience in configured sessions, with daily apartment monitoring across its current platform coverage.