TrackerMetricscarousel-rate

Carousel rate carousel-rate

Whether a run received a product carousel (0/1 per prompt × mode, from `shopping.products`). The daily median across all runs shows what share of the prompt set triggers shopping answers — so far the buying-advice categories (K2, K3, K7). If the rate shifts without a prompt-set change, OpenAI has changed the carousel threshold.

121 data points since 20 Aug 2026 · Download: JSON · CSV

Median across prompts, per mode

00.5120 Aug 2026: Baseline (extractor, prompt-set, account)instant 20 Aug 2026: 0instant 21 Aug 2026: 0instant 22 Aug 2026: 0instant 23 Aug 2026: 0instant 26 Aug 2026: 0thinking 20 Aug 2026: 0thinking 21 Aug 2026: 0thinking 22 Aug 2026: 0thinking 23 Aug 2026: 0thinking 26 Aug 2026: 020 Aug21 Aug22 Aug23 Aug24 Aug25 Aug26 Aug
instant thinking

Values per prompt

Prompt20 Aug 202621 Aug 202622 Aug 202623 Aug 202624 Aug 202625 Aug 202626 Aug 2026
K1-01 instant00000
K1-01 thinking00000
K1-02 instant000
K10-01 instant000
K11-01 instant000
K11-02 instant000
K12-01 instant000
K12-02 instant000
K13-01 instant000
K14-01 instant0000
K14-01 thinking0000
K14-02 instant0000
K14-02 thinking0000
K2-01 instant11111
K2-01 thinking11111
K2-02 instant111
K2-03 instant111
K3-01 instant111
K3-02 instant111
K4-01 instant00000
K4-01 thinking00000
K4-02 instant000
K5-01 instant00000
K5-01 thinking0000
K5-02 instant000
K6-01 instant00000
K6-01 thinking0000
K6-02 instant000
K7-01 instant000
K7-02 instant111
K8-01 instant000
K9-01 instant000
K9-02 instant000