Z.ai: GLM 5.3 Performance
Try GLM 5.3 on us. Starter credit covers your first requests, then one key reaches all 33 provider routes with automatic failover.
z-ai/glm-5.3
Measured performance
Continuously sampled p50/p95 time-to-first-token (TTFT), effective throughput, and success rate for Z.ai: GLM 5.3. 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 |
|---|---|---|---|---|---|---|
| crusoe | 816 ms | 816 ms | 111 tok/s n=1 | 100.00% | — | 1 |
| parasail | 867 ms | 938 ms | 17 tok/s n=1 | 100.00% | — | 2 |
| engy | 926 ms | 926 ms | 171 tok/s n=1 | 100.00% | — | 1 |
| tinfoil | 949 ms | 2636 ms | 88 tok/s n=1 | 100.00% | — | 4 |
| fireworks | 974 ms | 2240 ms | 178 tok/s n=1 | 100.00% | — | 2 |
| makora | 975 ms | 1396 ms | 236 tok/s n=1 | 100.00% | — | 3 |
| friendli | 990 ms | 1277 ms | 274 tok/s n=1 | 100.00% | — | 5 |
| nebius | 1000 ms | 1074 ms | 4 tok/s n=1 | 100.00% | — | 2 |
| baseten | 1087 ms | 2869 ms | 98 tok/s n=1 | 100.00% | — | 2 |
| io-net | 1101 ms | 1101 ms | 100 tok/s n=1 | 100.00% | — | 1 |
| sail-research | 1185 ms | 1185 ms | 111 tok/s n=1 | 100.00% | — | 1 |
| thinkingmachines | 1189 ms | 2396 ms | 159 tok/s n=2 | 100.00% | — | 8 |
| zro | 1332 ms | 1993 ms | 184 tok/s n=2 | 100.00% | — | 3 |
| akashml | 1628 ms | 2425 ms | 112 tok/s n=1 | 100.00% | — | 3 |
| pearl | 1645 ms | 2934 ms | 179 tok/s n=1 | 100.00% | — | 2 |
| nscale | 1701 ms | 1911 ms | 127 tok/s n=1 | 100.00% | — | 2 |
| zai | 1845 ms | 1845 ms | 81 tok/s n=1 | 100.00% | — | 1 |
| arcee | 1997 ms | 1997 ms | 81 tok/s n=1 | 100.00% | — | 1 |
| lyceum | 2205 ms | 2205 ms | 320 tok/s n=1 | 100.00% | — | 1 |
| streamlake | 2266 ms | 2323 ms | 77 tok/s n=1 | 100.00% | — | 2 |
| tencent | 2839 ms | 2980 ms | 57 tok/s n=1 | 100.00% | — | 2 |
| zero-g | 3009 ms | 4365 ms | 93 tok/s n=2 | 100.00% | — | 7 |
| cloudflare-workers-ai | 3703 ms | 3703 ms | 102 tok/s n=1 | 100.00% | — | 1 |
| atlas-cloud | — | — | 109 tok/s n=1 | — | — | 0 |
| boundless | — | — | 78 tok/s n=1 | — | — | 0 |
| featherless | — | — | 182 tok/s n=1 | — | — | 0 |
| inceptron | — | — | — | — | 5 provider_error |
0 |
| novita | — | — | 77 tok/s n=1 | — | — | 0 |
| privatemode | — | — | — | 0.00% | 2 provider_auth_config |
2 |
| siliconflow | — | — | 159 tok/s n=1 | — | — | 0 |
| telnyx | — | — | 248 tok/s n=1 | — | — | 0 |
Full provider & model leaderboard.
33 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.