llmcloud.ai
Platform · Reliability

Six nines out of ninety-fours.

Frontier providers hover around 94% single-region uptime. Stack three, route around outages, and combined availability crosses 99.9999% — the router does it per request, mid-stream.

94.0%
Typical single-provider uptime
99.64%
Two-provider failover
99.9999%
Three-provider automatic failover

How failover works

  1. 1. Router selects primary provider per the auto:* policy.
  2. 2. On 5xx / timeout / rate-limit, retry a healthy peer.
  3. 3. Mid-stream disconnect? Resume against a peer with the partial context replayed.
  4. 4. Provider health signals decay over 60s — recovering providers get traffic back.
  5. 5. Client sees a single response with x-llmcloud-provider attributing the winner.
Pin your own fallback chain
{
  "model": "auto:quality",
  "route": {
    "prefer": ["anthropic/claude-4.5-sonnet"],
    "fallback": [
      "openai/gpt-5",
      "google/gemini-3-pro",
      "fireworks/deepseek-v3.1"
    ],
    "on_error": ["timeout", "rate_limit", "5xx"],
    "max_retries": 3
  }
}

Built into every request

retries

Idempotent auto-retry

Timeouts, 429s and 5xx retry against a healthy peer with exponential backoff and a max-latency budget.

mid-stream

Mid-stream continuation

If a stream drops halfway, the router resumes on a peer and re-emits only the missing tokens — clients never see a broken stream.

health

Live provider scoring

Real-time p50 latency, error rate and rate-limit headroom feed the router. Sick providers get de-prioritized within seconds.

regions

Multi-region posture

us-east, eu-west and apac clusters serve traffic locally with cross-region failover as a last resort.

sla

99.99% managed SLA

Enterprise plans include a 99.99% router SLA with service credits — independent of any single upstream.

status

Public status page

Live incident feed and per-provider health at status.llmcloud.ai. Subscribe by webhook, email or RSS.

How does mid-stream failover not duplicate tokens?+

The router keeps a rolling window of what the client has already received. On resume it re-prompts the peer with the original messages plus an assistant continuation prefix, then dedupes the overlap before emitting deltas.

Does failover work with tool calls?+

Yes. Function-call schemas are provider-neutral. If a call is emitted before the failure, the router replays the tool_result on the peer and continues.

Can I disable failover for a request?+

Pass route.max_retries: 0 or set X-LLMCloud-Retry: none. Useful for evals where you want exact-provider attribution.