By Evan Vega
Redis says its LangCache semantic cache can “Save 90% on API costs,” but the service is still in preview with no published price, and Redis’s own formula puts savings below the hit rate. Novel Cognition checked the claims against Redis’s documentation, reviewed AWS’s benchmark and published research, and measured the similarity threshold on 243 support-question pairs, where “cancel” and “pause” scored closer than a true rewording.
Related: Frontier Watch