Live · updated every 30s
Health Feed
Real-time circuit breaker state for top inference providers.
4 healthy1 degraded1 down
Provider
State
Uptime 24h
p50
p95
p99
Error rate
Cost /1k
GPT-4o
OpenAI
Healthy
99.87%
312ms
720ms
1240ms
0.12%
$0.0050+1.2%
Claude 3.5 Sonnet
Anthropic
Healthy
99.94%
289ms
650ms
1100ms
0.06%
$0.0030—
Llama 3.1 405B
Together AI
Degraded
97.30%
891ms
2400ms
4100ms
2.70%
$0.0009+8.4%
Mixtral 8x7B
Groq
Healthy
99.99%
48ms
120ms
310ms
0.01%
$0.0003-2.1%
Gemini 1.5 Pro
Google
Down
82.40%
—
—
—
17.60%
$0.0035—
Mistral Large
Mistral AI
Healthy
99.50%
420ms
980ms
1800ms
0.50%
$0.0040+0.5%
Agent SDK
Query this feed in under 1ms from any agent. The SDK maintains a local cache via persistent WebSocket — no blocking REST round-trip before each spend decision.
npm install @inferenceflow/oracle-sdkView docs →Snapshot taken · 8s ago