Deployment Planner

Measured on Ethernet

Measured results

Headline results from Ardelva’s own test runs, each with its key condition. Synthetic workloads; results vary with model, workload and configuration. Full test reports are available under NDA.

Throughput · Ethernet fabric

10,046tok/s

MetaX C5502 nodes · 1P1D · DeepSeek V4.0 Flash

5,300tok/s

MetaX C550Single node · DeepSeek V4.0 Flash

1,142tok/s

RTX PRO 50004 GPUs · 16 concurrent · Qwen3.8-Flash-Next

808tok/s

RTX PRO 5000GLM 5.3 Flash

Model evaluation · average / 100

Claude Opus 5.595.5

Qwen3.8-Flash-Next (xhigh)94.7

GLM 5.3 Flash (high)93.7

Six execution-graded systems-programming tasks; best reasoning tier per model.

Artificial Analysis Intelligence Index

Claude Opus 5.5 (max)58

GLM-5.3-Flash42

Qwen3.8-Flash-Next40

DeepSeek V4.1 Flash (max)39

Source: Artificial Analysis Intelligence Index v4.3.2, snapshot 2026-09-25; max = highest reasoning setting.

Unit economics

≤1.2×

Cost per million output tokens vs. the DeepSeek API peak/off-peak average, at ~75% utilization.

Scenario estimate; details on request. Plug your own hardware prices, utilization and workload into the planner above.