AlibabaAPI Host Benchmarks
Qwen3.7 Max — API Providers & Speed Matrix
Compare live throughput, TTFT latency distributions, prompt cache pricing, and hardware stacks across cloud hosting endpoints.
Fastest Provider
2538 t/s
Lowest Latency (TTFT)
0.09s P50
Best Price Value
$1.13 / $3.75
Blended Pricing Workload Calculator
Switch between standard conversational ratios, agentic caching loops, and heavy RAG workloads.
| Provider & Host | Hardware Stack | Output Speed (TPS) | Latency TTFT (P50) | Input Price ($/1M) | Output Price ($/1M) | Cache Read ($/1M) | Blended Price ($/1M) | SLA |
|---|---|---|---|---|---|---|---|---|
AlibabaOfficial | Cloud TPU / Nvidia H100 | 203 t/s | 2.31s (3.23s p90) | $2.5 | $7.5 | $0.25 | $3.75 | 99.99% |
Groq Cloud | Groq LPU v2 (SRAM) | 853 t/s | 0.09s (0.12s p90) | $2.13 | $6.38 | $0.2 | $3.19 | 99.95% |
Cerebras Cloud | CS-3 Wafer-Scale Engine | 2538 t/s | 0.12s (0.16s p90) | $2.25 | $6.75 | — | $3.38 | 99.90% |
Together AI | Nvidia H100 SXM5 | 305 t/s | 1.85s (2.54s p90) | $1.63 | $5.25 | $0.15 | $2.54 | 99.95% |
DeepInfra | Nvidia H100 / L40S | 264 t/s | 2.08s (2.89s p90) | $1.13 | $3.75 | $0.125 | $1.78 | 99.90% |
Fireworks AI | Nvidia B200 / H100 | 386 t/s | 1.62s (2.19s p90) | $1.5 | $4.88 | $0.15 | $2.34 | 99.99% |