MetaAPI Host Benchmarks
Muse Spark 1.1 — API Providers & Speed Matrix
Compare live throughput, TTFT latency distributions, prompt cache pricing, and hardware stacks across cloud hosting endpoints.
Fastest Provider
2600 t/s
Lowest Latency (TTFT)
0.09s P50
Best Price Value
$0.56 / $2.13
Blended Pricing Workload Calculator
Switch between standard conversational ratios, agentic caching loops, and heavy RAG workloads.
| Provider & Host | Hardware Stack | Output Speed (TPS) | Latency TTFT (P50) | Input Price ($/1M) | Output Price ($/1M) | Cache Read ($/1M) | Blended Price ($/1M) | SLA |
|---|---|---|---|---|---|---|---|---|
MetaOfficial | Cloud TPU / Nvidia H100 | 208 t/s | 2.47s (3.46s p90) | $1.25 | $4.25 | $0.125 | $2 | 99.99% |
Groq Cloud | Groq LPU v2 (SRAM) | 874 t/s | 0.09s (0.12s p90) | $1.06 | $3.61 | $0.1 | $1.7 | 99.95% |
Cerebras Cloud | CS-3 Wafer-Scale Engine | 2600 t/s | 0.12s (0.16s p90) | $1.13 | $3.83 | — | $1.8 | 99.90% |
Together AI | Nvidia H100 SXM5 | 312 t/s | 1.98s (2.72s p90) | $0.81 | $2.97 | $0.075 | $1.35 | 99.95% |
DeepInfra | Nvidia H100 / L40S | 270 t/s | 2.22s (3.09s p90) | $0.56 | $2.13 | $0.063 | $0.95 | 99.90% |
Fireworks AI | Nvidia B200 / H100 | 395 t/s | 1.73s (2.35s p90) | $0.75 | $2.76 | $0.075 | $1.25 | 99.99% |