Moonshot AIAPI Host Benchmarks
Kimi K2.7 Code — API Providers & Speed Matrix
Compare live throughput, TTFT latency distributions, prompt cache pricing, and hardware stacks across cloud hosting endpoints.
Fastest Provider
488 t/s
Lowest Latency (TTFT)
0.09s P50
Best Price Value
$0.43 / $2
Blended Pricing Workload Calculator
Switch between standard conversational ratios, agentic caching loops, and heavy RAG workloads.
| Provider & Host | Hardware Stack | Output Speed (TPS) | Latency TTFT (P50) | Input Price ($/1M) | Output Price ($/1M) | Cache Read ($/1M) | Blended Price ($/1M) | SLA |
|---|---|---|---|---|---|---|---|---|
Moonshot AIOfficial | Cloud TPU / Nvidia H100 | 39 t/s | 2.87s (4.02s p90) | $0.95 | $4 | $0.095 | $1.71 | 99.99% |
Groq Cloud | Groq LPU v2 (SRAM) | 164 t/s | 0.09s (0.12s p90) | $0.81 | $3.4 | $0.076 | $1.46 | 99.95% |
Cerebras Cloud | CS-3 Wafer-Scale Engine | 488 t/s | 0.12s (0.16s p90) | $0.85 | $3.6 | — | $1.54 | 99.90% |
Together AI | Nvidia H100 SXM5 | 59 t/s | 2.3s (3.16s p90) | $0.62 | $2.8 | $0.057 | $1.17 | 99.95% |
DeepInfra | Nvidia H100 / L40S | 51 t/s | 2.58s (3.59s p90) | $0.43 | $2 | $0.048 | $0.82 | 99.90% |
Fireworks AI | Nvidia B200 / H100 | 74 t/s | 2.01s (2.73s p90) | $0.57 | $2.6 | $0.057 | $1.08 | 99.99% |