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% · 24h141ms8m agogemini-2-5-pro0.0% · 24h133ms8m agoclaude-haiku-4-50.0% · 24h218ms8m agoclaude-sonnet-4-60.0% · 24h226ms8m agoclaude-sonnet-4-60.0% · 24h1006ms8m agoclaude-haiku-4-5-202510010.0% · 24h1090ms8m agogemini-3-flash-preview0.0% · 24h1432ms8m agogemini-2.5-flash100% · 24h3544ms8m agogemini-2.5-pro97.9% · 24h3631ms8m agoglm-4.60.0% · 24h1052ms8m agoqwen3.5-flash0.0% · 24h1090ms8m agogpt-5-nano57.3% · 24h1712ms8m agogpt-5.565.6% · 24h2318ms8m agoDeepSeekoperational · inference liveness
DeepSeek V4 Flash94.8% · 24h482ms8m agoDeepSeek V4 Pro94.8% · 24h430ms8m agoDeepInfradegraded · inference liveness
DeepSeek V4 Flash0.0% · 24h257ms8m agoMiniMax M2.50.0% · 24h265ms8m agoKimi K2.50.0% · 24h265ms8m agoOpenRouteroperational · inference liveness
openai/gpt-5.1100% · 24h1288ms8m agoopenai/gpt-4o100% · 24h735ms8m agoopenai/gpt-4o-mini100% · 24h1015ms8m agoanthropic/claude-haiku-4.5100% · 24h843ms8m agogoogle/gemini-2.5-flash-lite100% · 24h604ms8m agoGET /v1/models100% · 24h589ms8m agoRunwareoperational · reachable liveness
GET /v1/models100% · 24h496ms8m agofal.aioperational · reachable liveness
GET /v1/models100% · 24h500ms8m 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.