Router Core
Attested TLS, authorization, route candidates, provider fallback, and settlement/refund durability.
Check TrustedRouter availability across regions with live gateway, attestation, SDK, billing, provider fallback, latency, incident history, and router-core uptime signals.
Attested TLS, authorization, route candidates, provider fallback, and settlement/refund durability.
Dashboard, billing, key management, docs, and public status surfaces.
Fresh connections include DNS, TCP, and TLS setup. Reused connections show the request path after that setup is paid once.
| Connection | Endpoint | Region path | Total p50 | DNS | TCP | TLS | Gateway | Samples |
|---|---|---|---|---|---|---|---|---|
| Fresh connection | Global endpoint | eu-west-3 -> eu-west-3 | 224 ms | 26 ms | 21 ms | 113 ms | n/a | 3 |
| Reused connection | Global endpoint | eu-west-3 -> eu-west-3 | 17 ms | paid once | paid once | paid once | n/a | 3 |
| Fresh connection | Ireland gateway direct | eu-west-3 -> eu-west-1 | 295 ms | 5 ms | 132 ms | 33 ms | n/a | 3 |
| Reused connection | Ireland gateway direct | eu-west-3 -> eu-west-1 | 21 ms | paid once | paid once | paid once | n/a | 3 |
| Fresh connection | Paris gateway direct | eu-west-3 -> eu-west-3 | 236 ms | 6 ms | 96 ms | 94 ms | n/a | 3 |
| Reused connection | Paris gateway direct | eu-west-3 -> eu-west-3 | 14 ms | paid once | paid once | paid once | n/a | 3 |
Continuously measured per provider from the rotation probe and real traffic over the rolling benchmark set of up to 10,000 samples. Ranked by success rate, then p50 TTFT. Unsupported route and probe-configuration rows are shown separately, not counted as provider downtime. This is an informational provider watch — a flaky upstream model does NOT count against router-core uptime above. Full leaderboard.
| Provider | Success | p50 TTFT | Error rate | Top error | Config excluded | Samples |
|---|---|---|---|---|---|---|
| privatemode | 100.00% | 367 ms | 0% | — | 68 provider_auth_config |
16 |
| regolo | 100.00% | 403 ms | 0% | — | — | 62 |
| cerebras | 100.00% | 404 ms | 0% | — | — | 75 |
| hixtube | 100.00% | 480 ms | 0% | — | — | 70 |
| mistral | 100.00% | 502 ms | 0% | — | — | 83 |
| together | 100.00% | 505 ms | 0% | — | — | 71 |
| nextbit-eu | 100.00% | 519 ms | 0% | — | — | 70 |
| nextbit | 100.00% | 538 ms | 0% | — | — | 69 |
| lyceum | 100.00% | 605 ms | 0% | — | — | 64 |
| nscale | 100.00% | 662 ms | 0% | — | — | 66 |
| inception | 100.00% | 679 ms | 0% | — | — | 74 |
| akashml | 100.00% | 681 ms | 0% | — | — | 67 |
| alibaba | 100.00% | 699 ms | 0% | — | 5 probe_config_error |
45 |
| fireworks | 100.00% | 739 ms | 0% | — | — | 70 |
| reka | 100.00% | 839 ms | 0% | — | — | 78 |
| makora | 100.00% | 883 ms | 0% | — | — | 65 |
| mancer | 100.00% | 892 ms | 0% | — | — | 53 |
| coralbricks | 100.00% | 899 ms | 0% | — | — | 81 |
| aion-labs | 100.00% | 927 ms | 0% | — | — | 56 |
| deepinfra | 100.00% | 939 ms | 0% | — | — | 81 |
| minimax | 100.00% | 945 ms | 0% | — | — | 70 |
| google-ai-studio | 100.00% | 947 ms | 0% | — | — | 74 |
| upstage | 100.00% | 978 ms | 0% | — | — | 80 |
| wafer | 100.00% | 989 ms | 0% | — | 6 probe_config_error |
69 |
| anthropic | 100.00% | 994 ms | 0% | — | — | 63 |
| openai | 100.00% | 1020 ms | 0% | — | 4 probe_config_error |
56 |
| io-net | 100.00% | 1021 ms | 0% | — | — | 69 |
| abliteration | 100.00% | 1034 ms | 0% | — | — | 73 |
| deepseek | 100.00% | 1052 ms | 0% | — | — | 72 |
| boundless | 100.00% | 1056 ms | 0% | — | — | 61 |
| neurometric | 100.00% | 1112 ms | 0% | — | — | 82 |
| thinkingmachines | 100.00% | 1150 ms | 0% | — | — | 88 |
| zai | 100.00% | 1207 ms | 0% | — | — | 73 |
| byteplus | 100.00% | 1356 ms | 0% | — | — | 54 |
| confidential-ai | 100.00% | 1497 ms | 0% | — | — | 70 |
| tinfoil | 100.00% | 1604 ms | 0% | — | — | 73 |
| kimi | 100.00% | 1906 ms | 0% | — | — | 81 |
| azure | 100.00% | 1987 ms | 0% | — | — | 80 |
| tencent | 100.00% | 2000 ms | 0% | — | — | 61 |
| arcee | 100.00% | 2258 ms | 0% | — | — | 76 |
| siliconflow | 98.73% | 1181 ms | 1.3% | TimeoutError |
— | 79 |
| grok | 98.68% | 1237 ms | 1.3% | server_error |
— | 76 |
| scaledown | 98.61% | 834 ms | 1.4% | TimeoutError |
— | 72 |
| wandb | 98.61% | 1022 ms | 1.4% | TimeoutError |
— | 72 |
| amazon-bedrock | 98.53% | 3367 ms | 1.5% | provider_error |
— | 68 |
| nebius | 98.51% | 440 ms | 1.5% | TimeoutError |
— | 67 |
| gmi | 98.44% | 2715 ms | 1.6% | TimeoutError |
— | 64 |
| meta | 97.30% | 1552 ms | 2.7% | TimeoutError |
— | 74 |
| parasail | 97.26% | 709 ms | 2.7% | provider_error |
— | 73 |
| venice | 97.10% | 1006 ms | 2.9% | TimeoutError |
— | 69 |
| friendli | 96.55% | 754 ms | 3.5% | TimeoutError |
— | 58 |
| scaleway | 95.45% | 376 ms | 4.5% | TimeoutError |
— | 66 |
| featherless | 92.42% | 2061 ms | 7.6% | server_error |
1 unsupported_route |
66 |
| novita | 91.94% | 1392 ms | 8.1% | TimeoutError |
2 unsupported_route |
62 |
| zro | 91.76% | 674 ms | 8.2% | TimeoutError |
1 provider_auth_config |
85 |
| baseten | 90.41% | 757 ms | 9.6% | provider_error |
— | 73 |
| muna | 88.61% | 654 ms | 11.4% | rate_limit_error |
— | 79 |
| sail-research | 86.54% | 985 ms | 13.5% | TimeoutError |
— | 52 |
| phala | 85.94% | 2404 ms | 14.1% | TimeoutError |
— | 64 |
| nvidia-nim | 71.95% | 548 ms | 28.1% | TimeoutError |
6 provider_error |
82 |
| maple | 53.95% | 1527 ms | 46.1% | provider_error |
— | 76 |
| abliterate | 0.00% | — | 100.0% | provider_error |
— | 67 |
| atlas-cloud | 0.00% | — | 100.0% | provider_error |
— | 72 |
| chutes | 0.00% | — | 100.0% | TimeoutError |
1 unsupported_route |
70 |
| cloudflare-workers-ai | 0.00% | — | 100.0% | provider_error |
1 router_error |
55 |
| databricks | 0.00% | — | 100.0% | provider_error |
— | 70 |
| digitalocean | 0.00% | — | 100.0% | provider_error |
— | 70 |
| engy | 0.00% | — | 100.0% | provider_error |
— | 59 |
| google-vertex | 0.00% | — | 100.0% | provider_error |
— | 67 |
| greenference | 0.00% | — | 100.0% | provider_error |
— | 74 |
| inceptron | 0.00% | — | 100.0% | provider_error |
— | 74 |
| morph | 0.00% | — | 100.0% | provider_error |
— | 79 |
| pearl | 0.00% | — | 100.0% | provider_error |
— | 63 |
| relace | 0.00% | — | 100.0% | provider_error |
— | 78 |
| sambanova | 0.00% | — | 100.0% | provider_error |
77 provider_auth_config |
2 |
| stepfun | 0.00% | — | 100.0% | provider_error |
— | 65 |
| streamlake | 0.00% | — | 100.0% | provider_error |
— | 63 |
| telnyx | 0.00% | — | 100.0% | provider_error |
— | 51 |
| xiaomi | 0.00% | — | 100.0% | provider_error |
— | 57 |
| zero-g | 0.00% | — | 100.0% | provider_error |
— | 75 |
Each row is backed by live US/EU synthetic checks. Prompt and output content are not stored.
api.trustedrouter.com attested TLS reachability and trust checks.
Ireland load balancer addressed directly: attested TLS and measurement-pinned attestation from a healthy Ireland enclave behind it. Green means the region is serving, not that every Ireland enclave is healthy.
Paris load balancer addressed directly: attested TLS and measurement-pinned attestation from a healthy Paris enclave behind it. Green means the region is serving, not that every Paris enclave is healthy.
Nonce and digest verification for public attested gateways.
End-to-end model calls through the attested gateway: OpenAI SDK and Responses round-trips with verified output.
Precomputed Router Core rollups. Control Plane and upstream providers are measured separately.
| Window | Status | Samples | Groups |
|---|---|---|---|
| 5m | up | 6 | 2 |
| 24h | up | 1430 | 2 |
| 48h | degraded | 2868 | 2 |
| Month | Status | Uptime | Samples |
|---|---|---|---|
| 2026-11 | degraded | 99.9508% | 2034 |
| 2026-10 | degraded | 99.9753% | 44584 |
| 2026-09 | degraded | >99.99% | 43118 |
| 2026-08 | trust degraded | 95.0318% | 87034 |
Samples that were not operational in the last 24 hours.
2026-11-01T17:00:00Z · responses_pong · eu-west-3 · hourly rollup · 1 failed / 29 · ReadTimeout
Every cloud runs its own probes and builds this page from its own samples, so these are independent reports rather than copies of one. If the cloud serving this page is the thing that is down, the others keep answering. Every cloud also watches every other cloud's page each pass and flags one that stops reporting.