Model status
Measured reliability for the 252-model catalog across 27 model authors: uptime, generation speed, time to first token, and latency, from idapt production traffic. Pick a window; rows update live.
What these numbers are
Every figure is measured on idapt production inference traffic and refreshed hourly: uptime is the share of eligible requests that completed, speed is generation throughput in tokens per second, and TTFT is time to first token. This is not a lab benchmark and not a provider status page: traffic mix shifts what gets measured, and a model only appears once it clears a minimum sample threshold of distinct requests, users, and workspaces. No prompts or message content are ever stored.
Rank models by these figures on the speed and reliability boards, or embed a live badge from the badges and embeds page.