Zhipu AIAPI Host Benchmarks
GLM-5.3 — API Providers & Speed Matrix
Compare live throughput, TTFT latency distributions, prompt cache pricing, and hardware stacks across cloud hosting endpoints.
Fastest Provider
1125 t/s
Lowest Latency (TTFT)
0.09s P50
Best Price Value
$0.63 / $2.2
Blended Pricing Workload Calculator
Switch between standard conversational ratios, agentic caching loops, and heavy RAG workloads.
| Provider & Host | Hardware Stack | Output Speed (TPS) | Latency TTFT (P50) | Input Price ($/1M) | Output Price ($/1M) | Cache Read ($/1M) | Blended Price ($/1M) | SLA |
|---|---|---|---|---|---|---|---|---|
Zhipu AIOfficial | Cloud TPU / Nvidia H100 | 90 t/s | 1.57s (2.2s p90) | $1.4 | $4.4 | $0.14 | $2.15 | 99.99% |
Groq Cloud | Groq LPU v2 (SRAM) | 378 t/s | 0.09s (0.12s p90) | $1.19 | $3.74 | $0.112 | $1.83 | 99.95% |
Cerebras Cloud | CS-3 Wafer-Scale Engine | 1125 t/s | 0.12s (0.16s p90) | $1.26 | $3.96 | — | $1.94 | 99.90% |
Together AI | Nvidia H100 SXM5 | 135 t/s | 1.26s (1.73s p90) | $0.91 | $3.08 | $0.084 | $1.45 | 99.95% |
DeepInfra | Nvidia H100 / L40S | 117 t/s | 1.41s (1.96s p90) | $0.63 | $2.2 | $0.07 | $1.02 | 99.90% |
Fireworks AI | Nvidia B200 / H100 | 171 t/s | 1.1s (1.49s p90) | $0.84 | $2.86 | $0.084 | $1.34 | 99.99% |