Z.ai: GLM 5.2 Performance
Compare measured TTFT, TTFB, throughput, uptime, and route health for Z.ai: GLM 5.2 across TrustedRouter providers using metadata-only production probes.
z-ai/glm-5.2
Measured performance
Continuously sampled p50/p95 time-to-first-token (TTFT), time-to-first-byte (TTFB), effective throughput, and success rate for Z.ai: GLM 5.2. 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 | p50 TTFB | Effective throughput | Uptime | Config excluded | Availability samples |
|---|---|---|---|---|---|---|---|
| friendli | 1173 ms | 1587 ms | 1173 ms | 169 tok/s n=1 | 100.00% | — | 7 |
| wafer | 1218 ms | 8411 ms | 1218 ms | — | 100.00% | — | 14 |
| together | 1662 ms | 2782 ms | 1662 ms | — | 100.00% | — | 2 |
| engy | 1734 ms | 3411 ms | 1733 ms | 36 tok/s n=1 | 100.00% | — | 24 |
| deepinfra | 1777 ms | 27546 ms | 1777 ms | 34 tok/s n=1 | 100.00% | — | 2 |
| telnyx | 1797 ms | 2722 ms | 1797 ms | 157 tok/s n=1 | 100.00% | — | 4 |
| siliconflow | 1820 ms | 2917 ms | 1820 ms | 31 tok/s n=1 | 100.00% | — | 3 |
| makora | 1862 ms | 3720 ms | 1862 ms | — | 100.00% | — | 10 |
| phala | 2244 ms | 4392 ms | 2244 ms | 32 tok/s n=1 | 100.00% | — | 23 |
| fireworks | 2266 ms | 3516 ms | 2266 ms | 36 tok/s n=1 | 100.00% | — | 7 |
| digitalocean | 2558 ms | 3861 ms | 2557 ms | 69 tok/s n=1 | 100.00% | — | 5 |
| morph | 2836 ms | 4318 ms | 2835 ms | 28 tok/s n=1 | 100.00% | — | 11 |
| zero-g | 2881 ms | 5285 ms | 2880 ms | 50 tok/s n=1 | 100.00% | — | 20 |
| baseten | 2918 ms | 4034 ms | 2917 ms | 49 tok/s n=1 | 100.00% | — | 7 |
| novita | 3029 ms | 3029 ms | 3028 ms | 22 tok/s n=1 | 100.00% | — | 1 |
| venice | 3101 ms | 3666 ms | 3101 ms | 55 tok/s n=1 | 100.00% | — | 3 |
| zai | 3444 ms | 14580 ms | 2829 ms | 50 tok/s n=1 | 100.00% | — | 8 |
| atlas-cloud | 3453 ms | 3453 ms | 3453 ms | — | 100.00% | — | 1 |
| gmi | 3469 ms | 4535 ms | 3468 ms | 37 tok/s n=1 | 100.00% | — | 13 |
| crusoe | 3908 ms | 4417 ms | 3908 ms | — | 100.00% | — | 4 |
| inceptron | 4519 ms | 7975 ms | 4519 ms | — | 100.00% | — | 11 |
| parasail | 3597 ms | 7445 ms | 1327 ms | 29 tok/s n=1 | 97.50% | — | 40 |
| tinfoil | 5771 ms | 21440 ms | 5771 ms | — | 88.89% | — | 9 |
| chutes | 1438 ms | 10685 ms | 1438 ms | — | 33.33% | — | 6 |
| alibaba | — | — | — | 31 tok/s n=1 | — | — | 0 |
Full provider & model leaderboard.
45 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.