Start with a Hugging Face model, GPU, and real traffic shape. The result keeps fit, utilization, idle capacity, provider dates, and every caveat visible.
First
Check memory fit
Then
Price real utilization
Compare
API and GPU economics
Trust
Choose the better value
1 GPU, FP16, 50,000 requests/day. 500 input and 500 output tokens per request.
Estimated memory fit
No measured aggregate serving capacity matches this configuration. Memory fit and dated hourly rates remain available; monthly rent, replica counts, and API comparisons are withheld.
A matching API price and serving cost are required before this assumption changes a comparison.
Model
Qwen/Qwen2.5-0.5B-Instruct
Hardware
1 x NVIDIA H200 141 GB, FP16
Weights and KV cache stay separate so the fit is reproducible.
Single-stream physics only. Batched serving capacity is not established.
4,857 output tok/sCeiling
Ceiling, not an expected throughput result.
Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
Dated GPU-hour rental prices for this hardware. They are not monthly serving totals because throughput evidence is missing.
Each bar compares the labelled value on the same scale.
DigitalOcean, GPU Droplets: $4.47, on demand GPU-hour rate, captured 2026-08-08 (1 month ago). Hyperstack: $3.99, on demand GPU-hour rate, captured 2026-08-08 (1 month ago). Nebius: $4.50, on demand GPU-hour rate, captured 2026-08-08 (1 month ago). Cheapest: RunPod, Community Cloud: $3.59, on demand GPU-hour rate, captured 2026-08-08 (1 month ago). RunPod, Secure Cloud: $4.39, on demand GPU-hour rate, captured 2026-08-08 (1 month ago). Verda: $4.00, on demand GPU-hour rate, captured 2026-08-08 (1 month ago).
These are GPU-hour rental prices only. A monthly serving total requires throughput evidence and remains absent.
This line starts the selected vLLM shape. It is not a throughput reproduction claim.
vllm serve 'Qwen/Qwen2.5-0.5B-Instruct' --dtype float16 --tensor-parallel-size 1 --max-model-len 8192 --gpu-memory-utilization 0.9No site-wide multiplier is applied.
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
RunInfra Engine feasibility runtime-overhead policy
As of 2026-08-09
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
As of 2026-08-08
Embed code appears only when the calculator can publish a sourced cost for a fitting configuration.
The calculator does not have a complete throughput source for this workload, so it withheld the cost and does not generate embed code.
Model
Qwen/Qwen2.5-0.5B-Instruct
GPU
NVIDIA H200 141 GB
Requests/day
50,000
GPU count
1
Quantization
FP16
Tokens/request
500 in, 500 out