Neural Architecture Search Cost Calculator

Work out neural architecture search cost instantly with clear inputs, formula shown and shareable results.

About Neural Architecture Search Cost Calculator

Neural architecture search cost is candidates times training cost per candidate, and parallelism buys wall-clock time without reducing the bill. This is why proxy strategies dominate practice: training candidates for a few epochs, using weight sharing as in ENAS, or predicting final accuracy from early curves cuts hours per candidate by one to two orders of magnitude.

Formula

Search cost

GPU hours = candidates x hours per candidate; cost = GPU hours x hourly price; wall clock = GPU hours / parallel GPUs

Worked example

With the default values below, Neural Architecture Search Cost Calculator returns Total search cost of $4,200.00. Change any input above and every figure here is recalculated in your browser.

Inputs used

Candidate architectures evaluatedminimum 1: 500
GPU hours per candidateminimum 0.01 h: 3 h
GPU hourly costminimum 0: 2.8
GPUs running in parallelminimum 1: 16

Results

Total search cost: $4,200.00
GPU hours consumed: 1,500.0 GPU-h
Wall-clock duration: 3.91 days
Cost per candidate evaluated: $8.40

Step by step

1Values used
2Search cost
3Total search cost= 4,200.00
4GPU hours consumed= 1,500.0 GPU-h
5Wall-clock duration= 3.91 days
6Cost per candidate evaluated= 8.40

Assumptions and limits

The result is derived only from the 4 inputs listed above. Anything not among them is not part of the calculation.
In particular, this calculator does not take inflation, tax or compounding frequency into account.
Figures are rounded for display to 2 decimal places, so totals built from several rounded parts can differ slightly from the unrounded arithmetic.

How these calculators are built and tested is described on the Methodology page. If a figure looks wrong, please report it.

Frequently Asked Questions

Is NAS worth it over a known good architecture?

Rarely for standard vision or language tasks, where published architectures are strong. It pays off when optimising for unusual hardware constraints such as a specific latency or memory budget.

How do I reduce the cost?

Use a reduced proxy task, early stopping on poor candidates, weight sharing, and a search space constrained by prior knowledge instead of an open-ended one.