Cutoff Data

Model speed & uptime index

How fast each major model answers, and how often it fails, measured hourly from an independent client over the last 7 days.

Last reading 9/24/2026, 1:17:04 AM UTC

Failing right now: GPT-5 Nano, GPT-5 Mini, GPT-5.4 Mini, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna, Gemini 2.5 Pro, Gemini 2.5 Flash Lite, Gemini 2.5 Flash, Gemini 3.1 Pro (preview), Gemini 3.1 Flash Lite, Gemini 3.5 Flash, Gemini 3.6 Flash
ModelProviderMedian95th pctUptimeProbesNow
GPT-5 Nanoopenai/gpt-5-nanoOpenAI0%77Failing
GPT-5 Miniopenai/gpt-5-miniOpenAI0%77Failing
GPT-5.4 Miniopenai/gpt-5.4-miniOpenAI0%77Failing
GPT-5.6 Solopenai/gpt-5.6-solOpenAI0%77Failing
GPT-5.6 Terraopenai/gpt-5.6-terraOpenAI0%77Failing
GPT-5.6 Lunaopenai/gpt-5.6-lunaOpenAI0%77Failing
Gemini 2.5 Progoogle/gemini-2.5-proGoogle0%77Failing
Gemini 2.5 Flash Litegoogle/gemini-2.5-flash-liteGoogle0%77Failing
Gemini 2.5 Flashgoogle/gemini-2.5-flashGoogle0%77Failing
Gemini 3.1 Pro (preview)google/gemini-3.1-pro-previewGoogle0%77Failing
Gemini 3.1 Flash Litegoogle/gemini-3.1-flash-liteGoogle0%77Failing
Gemini 3.5 Flashgoogle/gemini-3.5-flashGoogle0%77Failing
Gemini 3.6 Flashgoogle/gemini-3.6-flashGoogle0%76Failing

How this is measured

Once an hour we send every tracked model the same short, deterministic prompt and record wall-clock time from request to completed response, plus whether the call succeeded.

Median and 95th-percentile figures are computed across all successful probes in the trailing 7-day window. Uptime is successful probes divided by total probes.

These are single-client measurements over one network route, at short prompt lengths. They are a fair basis for comparing models against each other under identical conditions, not a substitute for a provider’s own status page.