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 |
|---|---|---|---|---|---|---|
| baseten | 1521 ms | 15655 ms | 34 tok/s n=1 | 100.00% | — | 36 |
| relace | 3942 ms | 24825 ms | — | 100.00% | — | 13 |
| kimi | 5239 ms | 9641 ms | 35 tok/s n=1 | 100.00% | — | 164 |
| together | 1129 ms | 1129 ms | 33 tok/s n=2 | 100.00% | — | 1 |
| venice | 1536 ms | 1536 ms | 38 tok/s n=2 | 100.00% | — | 1 |
| engy | 2006 ms | 2006 ms | 45 tok/s n=2 | 100.00% | — | 1 |
| telnyx | 2848 ms | 2848 ms | 97 tok/s n=2 | 100.00% | — | 1 |
| fireworks | 2915 ms | 4051 ms | 41 tok/s n=2 | 100.00% | — | 3 |
| phala | 3359 ms | 23965 ms | 22 tok/s n=2 | 85.71% | — | 7 |
| arcee | 3407 ms | 5809 ms | — | 100.00% | — | 5 |
| wafer | 5726 ms | 6928 ms | 48 tok/s n=2 | 100.00% | — | 2 |
| gmi | 5763 ms | 8418 ms | 45 tok/s n=2 | 100.00% | — | 8 |
| deepinfra | — | — | 25 tok/s n=2 | 100.00% | — | 1 |
| tinfoil | — | — | 26 tok/s n=2 | 100.00% | 3 probe_config_error |
1 |
| atlas-cloud | — | — | 24 tok/s n=2 | — | — | 0 |
| cloudflare-workers-ai | — | — | 25 tok/s n=2 | — | — | 0 |
| featherless | — | — | 24 tok/s n=2 | — | 5 probe_config_error |
0 |
| io-net | — | — | 42 tok/s n=2 | — | — | 0 |
| makora | — | — | 8 tok/s n=1 | — | 4 probe_config_error |
0 |
| nebius | — | — | 10 tok/s n=1 | 0.00% | — | 1 |
| novita | — | — | 11 tok/s n=1 | — | — | 0 |
| nvidia-nim | — | — | 3 tok/s n=1 | 0.00% | — | 2 |
| siliconflow | — | — | 28 tok/s n=1 | — | — | 0 |
Full provider & model leaderboard.
23 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.