Some providers degraded
Liveness pings every model every 15 minutes (a tiny max_tokens: 16 completion). A daily integrity suite then verifies tool calls, max_tokens, hidden-prompt injection, token over-counting vs a trusted baseline, and prompt caching — open a provider for the breakdown.
gemini-2-5-flash0.0% · 24h153ms38s agogemini-2-5-pro0.0% · 24h147ms38s agoclaude-haiku-4-50.0% · 24h188ms38s agoclaude-sonnet-4-60.0% · 24h179ms38s agoDeepSeekoperational · inference liveness
DeepSeek V4.1 Flash100% · 24h418ms38s agoDeepSeek V4 Pro99.0% · 24h442ms38s agoDeepInfradegraded · inference liveness
DeepSeek V4 Flash0.0% · 24h302ms38s agoMiniMax M2.50.0% · 24h276ms38s agoKimi K2.50.0% · 24h280ms38s agoOpenRouteroperational · inference liveness
openai/gpt-5.1100% · 24h532ms38s agoopenai/gpt-4o100% · 24h786ms38s agoopenai/gpt-4o-mini100% · 24h633ms38s agoanthropic/claude-haiku-4.5100% · 24h758ms38s agogoogle/gemini-2.5-flash-lite100% · 24h596ms38s agoGET /v1/models100% · 24h660ms38s agoRunwareoperational · reachable liveness
GET /v1/models100% · 24h537ms38s agofal.aioperational · reachable liveness
GET /v1/models100% · 24h607ms38s agoHistory kept 30 days · last 12h of liveness shown above. ai-lcr routes to the cheapest healthy provider, falling back automatically when one goes down.