[{"data":1,"prerenderedAt":30},["ShallowReactive",2],{"nr-en-two-ai-endpoints-measured-latency-change-frankfurt":3},{"slug":4,"title":5,"dek":6,"date":7,"time":8,"publishedAt":9,"updated":10,"updatedAt":10,"dateFmt":11,"updatedFmt":10,"kind":12,"tier":13,"author":14,"authorName":15,"topics":16,"tracker":22,"trackerLabel":23,"headlineStat":24,"image":25,"ogImage":26,"imageAlt":5,"csv":10,"minutes":27,"words":28,"html":29},"two-ai-endpoints-measured-latency-change-frankfurt","2 AI endpoints with measured latency change from Frankfurt","The i6eal AI Latency Index measures a slight reduction in time to first content token for two endpoints. The measurement applies only to each route and is not a statement about providers in general.","2026-10-12","08:46","2026-10-12T08:46:00+02:00","","October 12, 2026","daten","standard","ideal-syka","Ideal Syka",[17,18,19,20,21],"Latency","Frankfurt","Gateway","OpenAI","Anthropic","\u002Fki-latenz","To the AI Latency Index","GPT-5 mini: 743 ms → 658 ms (−11.44 %); Claude Haiku 4.5: 776 ms → 693 ms (−10.70 %)","\u002Fnewsroom\u002Fimg\u002Ftwo-ai-endpoints-measured-latency-change-frankfurt.webp","\u002Fog-nr\u002Ftwo-ai-endpoints-measured-latency-change-frankfurt.en.png",2,377,"\u003Ch2>What was measured\u003C\u002Fh2>\n\u003Cp>The i6eal AI Latency Index measures time to first content token (TTFT) from a Lambda function in Frankfurt. Two completed UTC weeks are compared: 28 September to 4 October 2026 and 5 to 11 October 2026. A change is observed for two endpoints: GPT-5 mini (OpenAI) and Claude Haiku 4.5 (Anthropic), each via the gateway route. Both endpoints have a paired value on all seven weekdays.\u003C\u002Fp>\n\u003Ch2>How to read the numbers\u003C\u002Fh2>\n\u003Cp>The changes always refer only to the same endpoint against its own prior week. They do not allow a comparison between the two models, nor between the direct route and the gateway route. The index also assesses only a short synthetic request, not a complete workflow.\u003C\u002Fp>\n\u003Ch2>What the measurement does not show\u003C\u002Fh2>\n\u003Cp>The observed differences establish neither a cause nor statistical significance. The measurement says nothing about model quality, availability or reliability. If you are planning latency for a specific application, it is worth testing with your own prompts and your own time window.\u003C\u002Fp>\n\u003Ch2>Observed latency changes\u003C\u002Fh2>\n\u003Cdiv class=\"tbl-scroll\">\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Endpoint\u003C\u002Fth>\n\u003Cth>Route\u003C\u002Fth>\n\u003Cth align=\"right\">Paired days per week\u003C\u002Fth>\n\u003Cth align=\"right\">Before TTFT (ms)\u003C\u002Fth>\n\u003Cth align=\"right\">Current TTFT (ms)\u003C\u002Fth>\n\u003Cth align=\"right\">Change\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>GPT-5 mini (OpenAI)\u003C\u002Ftd>\n\u003Ctd>gateway\u003C\u002Ftd>\n\u003Ctd align=\"right\">7\u003C\u002Ftd>\n\u003Ctd align=\"right\">743.00\u003C\u002Ftd>\n\u003Ctd align=\"right\">658.00\u003C\u002Ftd>\n\u003Ctd align=\"right\">-11.44 %\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>Claude Haiku 4.5 (Anthropic)\u003C\u002Ftd>\n\u003Ctd>gateway\u003C\u002Ftd>\n\u003Ctd align=\"right\">7\u003C\u002Ftd>\n\u003Ctd align=\"right\">776.00\u003C\u002Ftd>\n\u003Ctd align=\"right\">693.00\u003C\u002Ftd>\n\u003Ctd align=\"right\">-10.70 %\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Cp>2026-09-28–2026-10-04 → 2026-10-05–2026-10-11 (UTC).\u003C\u002Fp>\n\u003Ch2>Coverage and measurement limits\u003C\u002Fh2>\n\u003Cp>7 endpoints have sufficient paired days and unchanged model\u002Froute descriptions. 0 endpoints have insufficient paired days; 0 were excluded because model\u002Froute descriptions changed.\u003C\u002Fp>\n\u003Cp>TTFT measures time to the first content token from a Lambda in Frankfurt. This compares medians of daily measurements in two completed UTC weeks. Only weekdays with a value in both weeks count; each endpoint needs at least five paired days. The collector discards a warm-up and keeps the faster of two successful probes per measurement round. This is a short synthetic probe, not a measurement of your entire workload. Direct and gateway routes are not ranked against each other. Observed differences establish no cause, statistical significance, model quality or reliability.\u003C\u002Fp>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fi6eal.de\u002Fki-latenz\u002F\">i6eal KI-Latenz-Index\u003C\u002Fa>. Source timestamp: 2026-10-12T01:20:59.125Z.\u003C\u002Fp>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fi6eal.de\u002Fki-latenz\u002F\">i6eal\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cem>Editorially owned by \u003Ca href=\"\u002Fen\u002Fautor\u002Fideal-syka\">Ideal Syka\u003C\u002Fa>. Sources and method: \u003Ca href=\"\u002Fen\u002Fredaktion\">Newsroom & method\u003C\u002Fa>. Tips and corrections: \u003Ca href=\"mailto:ai@i6eal.de\">ai@i6eal.de\u003C\u002Fa>.\u003C\u002Fem>\u003C\u002Fp>\n",1791787700152]