MoonshotAI: Kimi K3 Performance
Compare measured TTFT, throughput, uptime, and route health for MoonshotAI: Kimi K3 across TrustedRouter providers using metadata-only production probes.
moonshotai/kimi-k3
Measured performance
Continuously sampled p50/p95 time-to-first-token (TTFT), effective throughput, and success rate for MoonshotAI: Kimi K3. Effective throughput uses provider-reported output tokens over complete request time. Unsupported route and probe-configuration rows are separated from provider downtime, and no prompt or output content is stored.
| Provider | p50 TTFT | p95 TTFT | Effective throughput | Uptime | Config excluded | Availability samples |
|---|---|---|---|---|---|---|
| relace | 3722 ms | 8967 ms | — | 100.00% | — | 11 |
| kimi | 5055 ms | 9048 ms | — | 100.00% | — | 94 |
| telnyx | 861 ms | 3246 ms | 111 tok/s n=1 | 100.00% | — | 2 |
| wafer | 891 ms | 1463 ms | 110 tok/s n=1 | 100.00% | — | 2 |
| together | 1585 ms | 1958 ms | 101 tok/s n=1 | 100.00% | — | 3 |
| venice | 2002 ms | 2002 ms | 76 tok/s n=1 | 100.00% | — | 1 |
| io-net | 2147 ms | 2147 ms | 150 tok/s n=2 | 100.00% | — | 1 |
| nebius | 2443 ms | 5676 ms | — | 100.00% | — | 3 |
| phala | 2545 ms | 3251 ms | — | 100.00% | — | 5 |
| arcee | 3334 ms | 6347 ms | 40 tok/s n=1 | 100.00% | — | 7 |
| sail-research | 3538 ms | 4139 ms | — | 100.00% | — | 7 |
| engy | 3590 ms | 3886 ms | 109 tok/s n=1 | 100.00% | — | 5 |
| featherless | 4010 ms | 4823 ms | — | 100.00% | — | 3 |
| makora | 4066 ms | 4066 ms | — | 33.33% | — | 3 |
| fireworks | 4072 ms | 5415 ms | 54 tok/s n=1 | 100.00% | — | 2 |
| siliconflow | 5109 ms | 13337 ms | 40 tok/s n=1 | 100.00% | — | 4 |
| cloudflare-workers-ai | 5457 ms | 6015 ms | — | 100.00% | — | 2 |
| gmi | 5523 ms | 13341 ms | 23 tok/s n=1 | 100.00% | — | 8 |
| tinfoil | 6737 ms | 9326 ms | — | 100.00% | — | 6 |
| baseten | — | — | — | 100.00% | — | 30 |
| atlas-cloud | — | — | 104 tok/s n=1 | — | — | 0 |
| deepinfra | — | — | 23 tok/s n=1 | — | — | 0 |
| novita | — | — | 45 tok/s n=1 | — | — | 0 |
| nvidia-nim | — | — | — | 0.00% | — | 3 |
Full provider & model leaderboard.
24 routes.
More routes give the auto router more room to fail over around provider 429 and 5xx responses.
Gateway overhead is measured separately.
Public status separates TLS/health overhead from full model latency so slow LLMs do not inflate the router metric.
Metadata rollups.
Status samples store latency, outcome, provider, model, route, cost, and region metadata only.
View public status or inspect provider routes.