RunInfraby RightNow
  • CatalogNew
  • Pricing
  • Research
  • Contact
DashboardSign inGet started

See what this model actually costs to serve.

Loading the fit, utilization assumptions, and cited price data.

See what this model actually costs to serve.

Fit is always evaluated when model sizing is available. GPU rent, paid idle capacity, API crossover, and measured optimization appear only when their required evidence exists. When the API wins, the page says so.

Model

Qwen/Qwen2.5-7B-Instruct

GPU

NVIDIA H100 SXM 80 GB

Requests/day

50,000

GPU count

1

Quantization

FP16

Tokens/request

500 in, 500 out

Charts appear only when evidence existsUtilization defaults to 30%

Configure the workload

Change the inputs, then calculate a shareable result.

Hugging Face model

Utilization assumption

Defaulted to 30% so the page does not manufacture a self-hosting win.

Derived utilization

66%

Idle share paid

34%

Utilization assumption30%

A steady production service with real peaks and troughs.

At this level about 70% of every rented GPU hour sits idle, and it is billed at the same rate as a busy one.

Priced configuration

The serving cost is calculated.

No comparable dated API rate is available for this model. The self-host bill remains reproducible without inventing an API baseline.

Model

Qwen/Qwen2.5-7B-Instruct

Hardware

1 x NVIDIA H100 SXM 80 GB, FP16

Memory fit

Weights and KV cache stay separate so the fit is reproducible.

15.2 GB
Model weights
14.6 GB
KV cache
29.9 GB
Total required
72 GB
Available

Reproduce with vLLM

The exact command attached to this result.

vllm serve 'Qwen/Qwen2.5-7B-Instruct' --dtype float16 --tensor-parallel-size 1 --max-model-len 32768 --gpu-memory-utilization 0.9

Priced operating point

Utilization and idle share are shown before the token price.

Paid idle capacity

34%

496.7 of 1,461 billed GPU-hours sit idle. The provider still charges every hour.

Busy GPU-hours

964.3

Useful work and paid idle time sum to the billed total above.

66%
Derived utilization
34%
Idle share paid
220 output tok/s
Throughput at concurrency 1
Physics ceiling

CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.

Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, captured 2026-08-08

$3,930.09
Monthly GPU rent

At full saturation

$3.40 / 1M output tokens

At actual utilization

$5.16 / 1M output tokens

See where the money moves

Each available view stays on its own evidence-backed scale. Missing rows remain visibly absent.

Cost vs volume

Each marker is a full engine recomputation. Self-host prices are never connected or interpolated.

Self-host, discrete pointsCurrent traffic

Scroll horizontally to read every volume point.

500 requests per day, 7,609,375 output tokens per month: self-host $1,965.05, no comparable API bill. 1,500 requests per day, 22,828,125 output tokens per month: self-host $1,965.05, no comparable API bill. 5,000 requests per day, 76,093,750 output tokens per month: self-host $1,965.05, no comparable API bill. 15,000 requests per day, 228,281,250 output tokens per month: self-host $1,965.05, no comparable API bill. 37,499 requests per day, 570,688,443 output tokens per month: self-host $1,965.05, no comparable API bill. 38,257 requests per day, 582,217,502 output tokens per month: self-host $3,930.09, no comparable API bill. 50,000 requests per day, 760,937,500 output tokens per month: self-host $3,930.09, no comparable API bill. 150,000 requests per day, 2,282,812,500 output tokens per month: self-host $7,860.18, no comparable API bill. 500,000 requests per day, 7,609,375,000 output tokens per month: self-host $27,510.63, no comparable API bill. 1,500,000 requests per day, 22,828,125,000 output tokens per month: self-host $78,601.80, no comparable API bill. 5,000,000 requests per day, 76,093,750,000 output tokens per month: self-host $261,350.99, no comparable API bill.

$261.4K$0500 req/day50,000 req/day5,000,000 req/day
Current self-host bill
$3,930.09
Current cheapest API
No comparable API bill
Monthly output tokens
760,937,500
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Replica counts change in whole units. Isolated self-host markers preserve each emitted price without drawing a smooth, straight, or step transition at a volume the engine did not emit. Missing API values produce no marker.

Busy vs idle GPU-hours

1,461 billed GPU-hours, split by the engine into useful work and paid idle capacity.

Scroll horizontally to read every chart label and value.

Busy GPU-hours: 964.3 hrs, Engine-reported share doing token work. Idle GPU-hours billed: 496.7 hrs, Engine-reported share paid while idle.

  • Busy GPU-hours964.3 hrs
  • Idle GPU-hours billed496.7 hrs
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Busy and idle hours are emitted by the engine and sum to the billed GPU-hours. The provider charges both, so the idle share remains visible beside the monthly rent.

Provider price spread, NVIDIA H100 SXM 80 GB

USD per GPU-hour. Cheapest captured offers and the rate used in this calculation are marked separately.

Scroll horizontally to read every chart label and value.

DigitalOcean, GPU Droplets: $4.41/hr, on demand GPU rate, captured 2026-08-08. Hyperstack: $3.20/hr, on demand GPU rate, captured 2026-08-08. Lambda Cloud: $4.29/hr, on demand GPU rate, captured 2026-08-08. Nebius: $3.85/hr, on demand GPU rate, captured 2026-08-08. Cheapest + Used: RunPod, Community Cloud: $2.69/hr, on demand GPU rate, captured 2026-08-08, used by the engine for this calculation. RunPod, Secure Cloud: $2.99/hr, on demand GPU rate, captured 2026-08-08. Verda: $3.25/hr, on demand GPU rate, captured 2026-08-08.

  • DigitalOcean, GPU Droplets$4.41/hr
  • Hyperstack$3.20/hr
  • Lambda Cloud$4.29/hr
  • Nebius$3.85/hr
  • Cheapest + Used: RunPod, Community Cloud$2.69/hr
  • RunPod, Secure Cloud$2.99/hr
  • Verda$3.25/hr
Source:
  • DigitalOcean, GPU Droplets, on demand, as of 2026-08-08
  • Hyperstack, on demand, as of 2026-08-08
  • Lambda Cloud, on demand, as of 2026-08-08
  • Nebius, on demand, as of 2026-08-08
  • RunPod, Community Cloud, on demand, as of 2026-08-08
  • RunPod, Secure Cloud, on demand, as of 2026-08-08
  • Verda, on demand, as of 2026-08-08
+What this chart means

Every rate is for the selected GPU and keeps its provider, tier, source, and capture date. A price without a date is not included.

Not measured on this config yet

No site-wide multiplier is applied. Run the exact configuration to create a receipt before comparing its measured baseline with RunInfra.

Run this configuration

Sources and caveats

Every published number keeps its source and date.

Provenance

  • Hugging Face model API revision sha

    As of 2026-08-08

  • Hugging Face model API safetensors.total

    As of 2026-08-08

  • config.json:max_position_embeddings

    As of 2026-08-08

  • config.json:architectures[0]

    As of 2026-08-08

  • config.json:torch_dtype

    As of 2026-08-08

  • NVIDIA H100 Tensor Core GPU datasheet

    As of 2026-08-08

  • RunInfra Engine parameter-only KV-cache heuristic

    As of 2026-08-08

  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters

    As of 2026-08-08

  • vLLM serve CLI documentation

    As of 2026-08-08

  • RunPod, Community Cloud pricing

    As of 2026-08-08

  • NVIDIA L40S product specification

    As of 2026-08-08

  • NVIDIA A100 product specification

    As of 2026-08-08

Caveats

  • +Layer, KV-head, and head-dimension metadata are absent from CalcDataset. KV cache is a parameter-only model using the paired Engine heuristic, not an exact architecture calculation.
  • +CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.
  • +Compute prefill ceiling: 64932.36617426988 tokens/sec. Bandwidth decode ceiling: 219.94279745581812 tokens/sec.
  • +CalcInput has no GPU provider selector. The calculation uses the lowest supplied on-demand rate for H100: RunPod, Community Cloud at $2.69/GPU-hour.
  • +Replica count covers the configured average daily token work. Peak concurrency is absent from CalcInput, so additional replicas needed for burst traffic are not priced.
  • +The utilization control is a break-even capacity scenario. derivedUtilization is separately computed from the configured daily token work.

Priced configuration

The serving cost is calculated.

No comparable dated API rate is available for this model. The self-host bill remains reproducible without inventing an API baseline.

Model

Qwen/Qwen2.5-7B-Instruct

Hardware

1 x NVIDIA H100 SXM 80 GB, FP16

Memory fit

Weights and KV cache stay separate so the fit is reproducible.

15.2 GB
Model weights
14.6 GB
KV cache
29.9 GB
Total required
72 GB
Available

Reproduce with vLLM

The exact command attached to this result.

vllm serve 'Qwen/Qwen2.5-7B-Instruct' --dtype float16 --tensor-parallel-size 1 --max-model-len 32768 --gpu-memory-utilization 0.9

Priced operating point

Utilization and idle share are shown before the token price.

Paid idle capacity

34%

496.7 of 1,461 billed GPU-hours sit idle. The provider still charges every hour.

Busy GPU-hours

964.3

Useful work and paid idle time sum to the billed total above.

66%
Derived utilization
34%
Idle share paid
220 output tok/s
Throughput at concurrency 1
Physics ceiling

CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.

Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, captured 2026-08-08

$3,930.09
Monthly GPU rent

At full saturation

$3.40 / 1M output tokens

At actual utilization

$5.16 / 1M output tokens

See where the money moves

Each available view stays on its own evidence-backed scale. Missing rows remain visibly absent.

Cost vs volume

Each marker is a full engine recomputation. Self-host prices are never connected or interpolated.

Self-host, discrete pointsCurrent traffic

Scroll horizontally to read every volume point.

500 requests per day, 7,609,375 output tokens per month: self-host $1,965.05, no comparable API bill. 1,500 requests per day, 22,828,125 output tokens per month: self-host $1,965.05, no comparable API bill. 5,000 requests per day, 76,093,750 output tokens per month: self-host $1,965.05, no comparable API bill. 15,000 requests per day, 228,281,250 output tokens per month: self-host $1,965.05, no comparable API bill. 37,499 requests per day, 570,688,443 output tokens per month: self-host $1,965.05, no comparable API bill. 38,257 requests per day, 582,217,502 output tokens per month: self-host $3,930.09, no comparable API bill. 50,000 requests per day, 760,937,500 output tokens per month: self-host $3,930.09, no comparable API bill. 150,000 requests per day, 2,282,812,500 output tokens per month: self-host $7,860.18, no comparable API bill. 500,000 requests per day, 7,609,375,000 output tokens per month: self-host $27,510.63, no comparable API bill. 1,500,000 requests per day, 22,828,125,000 output tokens per month: self-host $78,601.80, no comparable API bill. 5,000,000 requests per day, 76,093,750,000 output tokens per month: self-host $261,350.99, no comparable API bill.

$261.4K$0500 req/day50,000 req/day5,000,000 req/day
Current self-host bill
$3,930.09
Current cheapest API
No comparable API bill
Monthly output tokens
760,937,500
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Replica counts change in whole units. Isolated self-host markers preserve each emitted price without drawing a smooth, straight, or step transition at a volume the engine did not emit. Missing API values produce no marker.

Busy vs idle GPU-hours

1,461 billed GPU-hours, split by the engine into useful work and paid idle capacity.

Scroll horizontally to read every chart label and value.

Busy GPU-hours: 964.3 hrs, Engine-reported share doing token work. Idle GPU-hours billed: 496.7 hrs, Engine-reported share paid while idle.

  • Busy GPU-hours964.3 hrs
  • Idle GPU-hours billed496.7 hrs
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Busy and idle hours are emitted by the engine and sum to the billed GPU-hours. The provider charges both, so the idle share remains visible beside the monthly rent.

Provider price spread, NVIDIA H100 SXM 80 GB

USD per GPU-hour. Cheapest captured offers and the rate used in this calculation are marked separately.

Scroll horizontally to read every chart label and value.

DigitalOcean, GPU Droplets: $4.41/hr, on demand GPU rate, captured 2026-08-08. Hyperstack: $3.20/hr, on demand GPU rate, captured 2026-08-08. Lambda Cloud: $4.29/hr, on demand GPU rate, captured 2026-08-08. Nebius: $3.85/hr, on demand GPU rate, captured 2026-08-08. Cheapest + Used: RunPod, Community Cloud: $2.69/hr, on demand GPU rate, captured 2026-08-08, used by the engine for this calculation. RunPod, Secure Cloud: $2.99/hr, on demand GPU rate, captured 2026-08-08. Verda: $3.25/hr, on demand GPU rate, captured 2026-08-08.

  • DigitalOcean, GPU Droplets$4.41/hr
  • Hyperstack$3.20/hr
  • Lambda Cloud$4.29/hr
  • Nebius$3.85/hr
  • Cheapest + Used: RunPod, Community Cloud$2.69/hr
  • RunPod, Secure Cloud$2.99/hr
  • Verda$3.25/hr
Source:
  • DigitalOcean, GPU Droplets, on demand, as of 2026-08-08
  • Hyperstack, on demand, as of 2026-08-08
  • Lambda Cloud, on demand, as of 2026-08-08
  • Nebius, on demand, as of 2026-08-08
  • RunPod, Community Cloud, on demand, as of 2026-08-08
  • RunPod, Secure Cloud, on demand, as of 2026-08-08
  • Verda, on demand, as of 2026-08-08
+What this chart means

Every rate is for the selected GPU and keeps its provider, tier, source, and capture date. A price without a date is not included.

Not measured on this config yet

No site-wide multiplier is applied. Run the exact configuration to create a receipt before comparing its measured baseline with RunInfra.

Run this configuration

Sources and caveats

Every published number keeps its source and date.

Provenance

  • Hugging Face model API revision sha

    As of 2026-08-08

  • Hugging Face model API safetensors.total

    As of 2026-08-08

  • config.json:max_position_embeddings

    As of 2026-08-08

  • config.json:architectures[0]

    As of 2026-08-08

  • config.json:torch_dtype

    As of 2026-08-08

  • NVIDIA H100 Tensor Core GPU datasheet

    As of 2026-08-08

  • RunInfra Engine parameter-only KV-cache heuristic

    As of 2026-08-08

  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters

    As of 2026-08-08

  • vLLM serve CLI documentation

    As of 2026-08-08

  • RunPod, Community Cloud pricing

    As of 2026-08-08

  • NVIDIA L40S product specification

    As of 2026-08-08

  • NVIDIA A100 product specification

    As of 2026-08-08

Caveats

  • +Layer, KV-head, and head-dimension metadata are absent from CalcDataset. KV cache is a parameter-only model using the paired Engine heuristic, not an exact architecture calculation.
  • +CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.
  • +Compute prefill ceiling: 64932.36617426988 tokens/sec. Bandwidth decode ceiling: 219.94279745581812 tokens/sec.
  • +CalcInput has no GPU provider selector. The calculation uses the lowest supplied on-demand rate for H100: RunPod, Community Cloud at $2.69/GPU-hour.
  • +Replica count covers the configured average daily token work. Peak concurrency is absent from CalcInput, so additional replicas needed for burst traffic are not priced.
  • +The utilization control is a break-even capacity scenario. derivedUtilization is separately computed from the configured daily token work.

Priced configuration

The serving cost is calculated.

No comparable dated API rate is available for this model. The self-host bill remains reproducible without inventing an API baseline.

Model

Qwen/Qwen2.5-7B-Instruct

Hardware

1 x NVIDIA H100 SXM 80 GB, FP16

Memory fit

Weights and KV cache stay separate so the fit is reproducible.

15.2 GB
Model weights
14.6 GB
KV cache
29.9 GB
Total required
72 GB
Available

Reproduce with vLLM

The exact command attached to this result.

vllm serve 'Qwen/Qwen2.5-7B-Instruct' --dtype float16 --tensor-parallel-size 1 --max-model-len 32768 --gpu-memory-utilization 0.9

Priced operating point

Utilization and idle share are shown before the token price.

Paid idle capacity

34%

496.7 of 1,461 billed GPU-hours sit idle. The provider still charges every hour.

Busy GPU-hours

964.3

Useful work and paid idle time sum to the billed total above.

66%
Derived utilization
34%
Idle share paid
220 output tok/s
Throughput at concurrency 1
Physics ceiling

CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.

Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, captured 2026-08-08

$3,930.09
Monthly GPU rent

At full saturation

$3.40 / 1M output tokens

At actual utilization

$5.16 / 1M output tokens

See where the money moves

Each available view stays on its own evidence-backed scale. Missing rows remain visibly absent.

Cost vs volume

Each marker is a full engine recomputation. Self-host prices are never connected or interpolated.

Self-host, discrete pointsCurrent traffic

Scroll horizontally to read every volume point.

500 requests per day, 7,609,375 output tokens per month: self-host $1,965.05, no comparable API bill. 1,500 requests per day, 22,828,125 output tokens per month: self-host $1,965.05, no comparable API bill. 5,000 requests per day, 76,093,750 output tokens per month: self-host $1,965.05, no comparable API bill. 15,000 requests per day, 228,281,250 output tokens per month: self-host $1,965.05, no comparable API bill. 37,499 requests per day, 570,688,443 output tokens per month: self-host $1,965.05, no comparable API bill. 38,257 requests per day, 582,217,502 output tokens per month: self-host $3,930.09, no comparable API bill. 50,000 requests per day, 760,937,500 output tokens per month: self-host $3,930.09, no comparable API bill. 150,000 requests per day, 2,282,812,500 output tokens per month: self-host $7,860.18, no comparable API bill. 500,000 requests per day, 7,609,375,000 output tokens per month: self-host $27,510.63, no comparable API bill. 1,500,000 requests per day, 22,828,125,000 output tokens per month: self-host $78,601.80, no comparable API bill. 5,000,000 requests per day, 76,093,750,000 output tokens per month: self-host $261,350.99, no comparable API bill.

$261.4K$0500 req/day50,000 req/day5,000,000 req/day
Current self-host bill
$3,930.09
Current cheapest API
No comparable API bill
Monthly output tokens
760,937,500
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Replica counts change in whole units. Isolated self-host markers preserve each emitted price without drawing a smooth, straight, or step transition at a volume the engine did not emit. Missing API values produce no marker.

Busy vs idle GPU-hours

1,461 billed GPU-hours, split by the engine into useful work and paid idle capacity.

Scroll horizontally to read every chart label and value.

Busy GPU-hours: 964.3 hrs, Engine-reported share doing token work. Idle GPU-hours billed: 496.7 hrs, Engine-reported share paid while idle.

  • Busy GPU-hours964.3 hrs
  • Idle GPU-hours billed496.7 hrs
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Busy and idle hours are emitted by the engine and sum to the billed GPU-hours. The provider charges both, so the idle share remains visible beside the monthly rent.

Provider price spread, NVIDIA H100 SXM 80 GB

USD per GPU-hour. Cheapest captured offers and the rate used in this calculation are marked separately.

Scroll horizontally to read every chart label and value.

DigitalOcean, GPU Droplets: $4.41/hr, on demand GPU rate, captured 2026-08-08. Hyperstack: $3.20/hr, on demand GPU rate, captured 2026-08-08. Lambda Cloud: $4.29/hr, on demand GPU rate, captured 2026-08-08. Nebius: $3.85/hr, on demand GPU rate, captured 2026-08-08. Cheapest + Used: RunPod, Community Cloud: $2.69/hr, on demand GPU rate, captured 2026-08-08, used by the engine for this calculation. RunPod, Secure Cloud: $2.99/hr, on demand GPU rate, captured 2026-08-08. Verda: $3.25/hr, on demand GPU rate, captured 2026-08-08.

  • DigitalOcean, GPU Droplets$4.41/hr
  • Hyperstack$3.20/hr
  • Lambda Cloud$4.29/hr
  • Nebius$3.85/hr
  • Cheapest + Used: RunPod, Community Cloud$2.69/hr
  • RunPod, Secure Cloud$2.99/hr
  • Verda$3.25/hr
Source:
  • DigitalOcean, GPU Droplets, on demand, as of 2026-08-08
  • Hyperstack, on demand, as of 2026-08-08
  • Lambda Cloud, on demand, as of 2026-08-08
  • Nebius, on demand, as of 2026-08-08
  • RunPod, Community Cloud, on demand, as of 2026-08-08
  • RunPod, Secure Cloud, on demand, as of 2026-08-08
  • Verda, on demand, as of 2026-08-08
+What this chart means

Every rate is for the selected GPU and keeps its provider, tier, source, and capture date. A price without a date is not included.

Not measured on this config yet

No site-wide multiplier is applied. Run the exact configuration to create a receipt before comparing its measured baseline with RunInfra.

Run this configuration

Sources and caveats

Every published number keeps its source and date.

Provenance

  • Hugging Face model API revision sha

    As of 2026-08-08

  • Hugging Face model API safetensors.total

    As of 2026-08-08

  • config.json:max_position_embeddings

    As of 2026-08-08

  • config.json:architectures[0]

    As of 2026-08-08

  • config.json:torch_dtype

    As of 2026-08-08

  • NVIDIA H100 Tensor Core GPU datasheet

    As of 2026-08-08

  • RunInfra Engine parameter-only KV-cache heuristic

    As of 2026-08-08

  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters

    As of 2026-08-08

  • vLLM serve CLI documentation

    As of 2026-08-08

  • RunPod, Community Cloud pricing

    As of 2026-08-08

  • NVIDIA L40S product specification

    As of 2026-08-08

  • NVIDIA A100 product specification

    As of 2026-08-08

Caveats

  • +Layer, KV-head, and head-dimension metadata are absent from CalcDataset. KV cache is a parameter-only model using the paired Engine heuristic, not an exact architecture calculation.
  • +CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.
  • +Compute prefill ceiling: 64932.36617426988 tokens/sec. Bandwidth decode ceiling: 219.94279745581812 tokens/sec.
  • +CalcInput has no GPU provider selector. The calculation uses the lowest supplied on-demand rate for H100: RunPod, Community Cloud at $2.69/GPU-hour.
  • +Replica count covers the configured average daily token work. Peak concurrency is absent from CalcInput, so additional replicas needed for burst traffic are not priced.
  • +The utilization control is a break-even capacity scenario. derivedUtilization is separately computed from the configured daily token work.

Priced configuration

The serving cost is calculated.

No comparable dated API rate is available for this model. The self-host bill remains reproducible without inventing an API baseline.

Model

Qwen/Qwen2.5-7B-Instruct

Hardware

1 x NVIDIA H100 SXM 80 GB, FP16

Memory fit

Weights and KV cache stay separate so the fit is reproducible.

15.2 GB
Model weights
14.6 GB
KV cache
29.9 GB
Total required
72 GB
Available

Reproduce with vLLM

The exact command attached to this result.

vllm serve 'Qwen/Qwen2.5-7B-Instruct' --dtype float16 --tensor-parallel-size 1 --max-model-len 32768 --gpu-memory-utilization 0.9

Priced operating point

Utilization and idle share are shown before the token price.

Paid idle capacity

34%

496.7 of 1,461 billed GPU-hours sit idle. The provider still charges every hour.

Busy GPU-hours

964.3

Useful work and paid idle time sum to the billed total above.

66%
Derived utilization
34%
Idle share paid
220 output tok/s
Throughput at concurrency 1
Physics ceiling

CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.

Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, captured 2026-08-08

$3,930.09
Monthly GPU rent

At full saturation

$3.40 / 1M output tokens

At actual utilization

$5.16 / 1M output tokens

See where the money moves

Each available view stays on its own evidence-backed scale. Missing rows remain visibly absent.

Cost vs volume

Each marker is a full engine recomputation. Self-host prices are never connected or interpolated.

Self-host, discrete pointsCurrent traffic

Scroll horizontally to read every volume point.

500 requests per day, 7,609,375 output tokens per month: self-host $1,965.05, no comparable API bill. 1,500 requests per day, 22,828,125 output tokens per month: self-host $1,965.05, no comparable API bill. 5,000 requests per day, 76,093,750 output tokens per month: self-host $1,965.05, no comparable API bill. 15,000 requests per day, 228,281,250 output tokens per month: self-host $1,965.05, no comparable API bill. 37,499 requests per day, 570,688,443 output tokens per month: self-host $1,965.05, no comparable API bill. 38,257 requests per day, 582,217,502 output tokens per month: self-host $3,930.09, no comparable API bill. 50,000 requests per day, 760,937,500 output tokens per month: self-host $3,930.09, no comparable API bill. 150,000 requests per day, 2,282,812,500 output tokens per month: self-host $7,860.18, no comparable API bill. 500,000 requests per day, 7,609,375,000 output tokens per month: self-host $27,510.63, no comparable API bill. 1,500,000 requests per day, 22,828,125,000 output tokens per month: self-host $78,601.80, no comparable API bill. 5,000,000 requests per day, 76,093,750,000 output tokens per month: self-host $261,350.99, no comparable API bill.

$261.4K$0500 req/day50,000 req/day5,000,000 req/day
Current self-host bill
$3,930.09
Current cheapest API
No comparable API bill
Monthly output tokens
760,937,500
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Replica counts change in whole units. Isolated self-host markers preserve each emitted price without drawing a smooth, straight, or step transition at a volume the engine did not emit. Missing API values produce no marker.

Busy vs idle GPU-hours

1,461 billed GPU-hours, split by the engine into useful work and paid idle capacity.

Scroll horizontally to read every chart label and value.

Busy GPU-hours: 964.3 hrs, Engine-reported share doing token work. Idle GPU-hours billed: 496.7 hrs, Engine-reported share paid while idle.

  • Busy GPU-hours964.3 hrs
  • Idle GPU-hours billed496.7 hrs
Source:
  • RunPod, Community Cloud pricing, as of 2026-08-08
  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters, as of 2026-08-08
+What this chart means

Busy and idle hours are emitted by the engine and sum to the billed GPU-hours. The provider charges both, so the idle share remains visible beside the monthly rent.

Provider price spread, NVIDIA H100 SXM 80 GB

USD per GPU-hour. Cheapest captured offers and the rate used in this calculation are marked separately.

Scroll horizontally to read every chart label and value.

DigitalOcean, GPU Droplets: $4.41/hr, on demand GPU rate, captured 2026-08-08. Hyperstack: $3.20/hr, on demand GPU rate, captured 2026-08-08. Lambda Cloud: $4.29/hr, on demand GPU rate, captured 2026-08-08. Nebius: $3.85/hr, on demand GPU rate, captured 2026-08-08. Cheapest + Used: RunPod, Community Cloud: $2.69/hr, on demand GPU rate, captured 2026-08-08, used by the engine for this calculation. RunPod, Secure Cloud: $2.99/hr, on demand GPU rate, captured 2026-08-08. Verda: $3.25/hr, on demand GPU rate, captured 2026-08-08.

  • DigitalOcean, GPU Droplets$4.41/hr
  • Hyperstack$3.20/hr
  • Lambda Cloud$4.29/hr
  • Nebius$3.85/hr
  • Cheapest + Used: RunPod, Community Cloud$2.69/hr
  • RunPod, Secure Cloud$2.99/hr
  • Verda$3.25/hr
Source:
  • DigitalOcean, GPU Droplets, on demand, as of 2026-08-08
  • Hyperstack, on demand, as of 2026-08-08
  • Lambda Cloud, on demand, as of 2026-08-08
  • Nebius, on demand, as of 2026-08-08
  • RunPod, Community Cloud, on demand, as of 2026-08-08
  • RunPod, Secure Cloud, on demand, as of 2026-08-08
  • Verda, on demand, as of 2026-08-08
+What this chart means

Every rate is for the selected GPU and keeps its provider, tier, source, and capture date. A price without a date is not included.

Not measured on this config yet

No site-wide multiplier is applied. Run the exact configuration to create a receipt before comparing its measured baseline with RunInfra.

Run this configuration

Sources and caveats

Every published number keeps its source and date.

Provenance

  • Hugging Face model API revision sha

    As of 2026-08-08

  • Hugging Face model API safetensors.total

    As of 2026-08-08

  • config.json:max_position_embeddings

    As of 2026-08-08

  • config.json:architectures[0]

    As of 2026-08-08

  • config.json:torch_dtype

    As of 2026-08-08

  • NVIDIA H100 Tensor Core GPU datasheet

    As of 2026-08-08

  • RunInfra Engine parameter-only KV-cache heuristic

    As of 2026-08-08

  • Decode and prefill ceilings derived from GPU memory bandwidth, dense compute, and model parameters

    As of 2026-08-08

  • vLLM serve CLI documentation

    As of 2026-08-08

  • RunPod, Community Cloud pricing

    As of 2026-08-08

  • NVIDIA L40S product specification

    As of 2026-08-08

  • NVIDIA A100 product specification

    As of 2026-08-08

Caveats

  • +Layer, KV-head, and head-dimension metadata are absent from CalcDataset. KV cache is a parameter-only model using the paired Engine heuristic, not an exact architecture calculation.
  • +CEILING at 100% unit utilisation, not expected throughput. Decode is bounded by memory bandwidth and prefill by dense compute. Ideal tensor-parallel scaling is assumed for both. Real decode and prefill throughput are lower.
  • +Compute prefill ceiling: 64932.36617426988 tokens/sec. Bandwidth decode ceiling: 219.94279745581812 tokens/sec.
  • +CalcInput has no GPU provider selector. The calculation uses the lowest supplied on-demand rate for H100: RunPod, Community Cloud at $2.69/GPU-hour.
  • +Replica count covers the configured average daily token work. Peak concurrency is absent from CalcInput, so additional replicas needed for burst traffic are not priced.
  • +The utilization control is a break-even capacity scenario. derivedUtilization is separately computed from the configured daily token work.

Share this calculation

The badge keeps its evidence label and links to the full result, sources, and caveats.

Calculated by RunInfra

HTML

<a href="https://runinfra.ai/calc/qwen-2.5-7b/h100/50k-req-day"><img src="https://runinfra.ai/api/calc/badge?hfId=Qwen%2FQwen2.5-7B-Instruct&amp;gpuId=H100&amp;gpuCount=1&amp;requestsPerDay=50000&amp;avgInputTokens=500&amp;avgOutputTokens=500&amp;utilization=0.3&amp;quantization=fp16" alt="Calculated by RunInfra"></a>

Markdown

[![Calculated by RunInfra](https://runinfra.ai/api/calc/badge?hfId=Qwen%2FQwen2.5-7B-Instruct&gpuId=H100&gpuCount=1&requestsPerDay=50000&avgInputTokens=500&avgOutputTokens=500&utilization=0.3&quantization=fp16)](https://runinfra.ai/calc/qwen-2.5-7b/h100/50k-req-day)

Use it from an agent or script

GET the current inputs, or inspect the descriptor for accepted parameters and response fields.

Endpoint
GET https://runinfra.ai/api/calchttps://runinfra.ai/api/calc?hfId=Qwen%2FQwen2.5-7B-Instruct&gpuId=H100&gpuCount=1&requestsPerDay=50000&avgInputTokens=500&avgOutputTokens=500&utilization=0.3&quantization=fp16
Canonical result
https://runinfra.ai/calc/qwen-2.5-7b/h100/50k-req-day
Descriptor
https://runinfra.ai/api/calc/schema

28 GPU ids, 7 quantizations

Turn this exact configuration into a deployment, describe what you need

Describe how you want to deploy this configuration...
ModelsAuto engineAuto GPU
End-to-end encryption
Isolated GPU infrastructure
No training on your data
SOC 2 Type II
RunInfraby RightNow

© 2026 RunInfra. All rights reserved.

System status
Pipeline BuilderModelsCost CalculatorPricingStartupsBenchmarksDocsResearchNewsContact
Backed by
YCombinator
AICPA Type II
SOC 2
NVIDIA Inception ProgramNVIDIA Inception Program
Ask AI about RunInfra
Part of RightNow
SecurityDPAAUPCookiesTermsPrivacy