DeepFrugal Early Preview Find the cheapest way to run any model

Plan comparison: a step-by-step tutorial

Plan comparison, step by step: pick two plans, set one monthly workload, read the cheaper plan and the quota split, then share it. With screenshots.

4 min read DeepFrugal
Step-by-step tutorial for the plan comparison tool

The plan comparison tool values one monthly workload against two subscription plans. This walkthrough shows every step, from an empty page to a shareable result. For the idea behind a plan's quota, read how AI subscription credits work.

Step 1: open the page

Open /tools/plan-comparison/. The page starts empty: no plan and no workload. No plan is favoured, so the first comparison is yours to define.

The plan comparison tool opens with two empty plan boxes and no workload.
Step 1 โ€” the tool opens empty.

Step 2: pick two plans

Choose Plan A and Plan B from the two lists. Plans are grouped by gateway. Only plans with a published quota appear, because the calculator needs a quota to value usage. The workload editor appears once both plans are chosen.

The two plan pickers, grouped by gateway.
Step 2 โ€” pick Plan A and Plan B from the grouped lists.

Step 3: set the monthly workload

Load a preset or build the workload under Model mix. Each row is one model variant. Set the monthly usage for each token type:

  • Input โ€” fresh prompt tokens, in millions per month.
  • Output โ€” generated tokens, in millions per month.
  • Cached read โ€” prompt tokens served from cache, when the model bills them.
  • Cached write โ€” tokens written into the cache, when the provider charges.

Click Add model to add a row, and use the arrows to reorder the mix. The order matters: each plan grants one shared budget, and the calculator spends it in the listed order.

A workload with several model rows and their monthly usage.
Step 3 โ€” build the workload model by model.

Step 4: read the verdict

The callout names the cheaper plan and the monthly gap. The two plan cards show the Quota used, what is Covered by the quota, what falls Above the quota, the Average cost per 1M tokens and the Monthly cost; the cheaper one carries a badge.

The verdict and the two plan cards for the same workload.
Step 4 โ€” read the verdict and the two plan cards.

Step 5: open the quota allocation

Open How the quota is allocated for the per-model breakdown. It shows each model's usage, the amount above its own quota, and its share of the plan budget. The order you list the models in decides which one the shared budget covers first.

The per-model quota allocation for both plans.
Step 5 โ€” open the quota allocation panel.

Step 6: read the chart

Plan comparison details draws one average-cost curve per plan as your usage grows. Each curve falls inside the quota, reaches the plan's effective rate at its quota limit and climbs again above it towards the overflow base. The listed rate is a shared horizontal reference, and the red line is the cheapest pay-as-you-go route. The markers sit on the cheaper plan for the mix: the quota usage break-even, the plan break-even and where it loses to the cheapest route. Open How the graph is calculated for the method.

The two cost curves with their quota limits and reference lines.
Step 6 โ€” the two cost curves with their reference lines.

Step 7: toggle promo prices

Open Options and turn Promo prices off to see every figure at the pre-promo list rates and quotas. The cheaper plan can change when a promo ends.

Promo prices off, showing the pre-promo list rates.
Step 7 โ€” switch promo prices off to see the pre-promo figures.

Step 8: choose the overflow base

Still under Options, the Overflow billed at pills set how usage above a plan's quota is billed. The result changes with the base:

  • Listed price โ€” each model's listed rate (the default).
  • Cheapest pay-as-you-go โ€” the cheapest metered rate found for each model.
  • The cheaper of the two โ€” whichever of the two is lower per model.

Overflow billed at the cheapest pay-as-you-go rate.
Step 8 โ€” bill the overflow at the cheapest pay-as-you-go rate.

Step 9: share the comparison

The page writes its state to the URL. Change the plans, the models or the usage, and the address bar follows. Click Share to copy a link that reopens the same comparison. Under the result, Compare these plans model by model on monthly quota โ†’ opens the plan head-to-head, and Open the plan finder on this mix โ†’ ranks every plan on the same workload.

The calculator is marked Experimental. Confirm every figure against the provider's own pricing before you rely on it.

Next step

Run your own comparison on the plan comparison tool. To weigh a plan against pay-as-you-go instead, open the break-even calculator, and use the live table to filter plans by provider, context, latency and privacy.

Frequently asked questions

What does the plan comparison tool do?

It values one monthly workload against two subscription plans and names the cheaper one, model by model, with the quota each plan spends.

Which plans can I compare?

Subscription plans that publish a quota. Both plans must carry every model in the workload.

What happens above a plan's quota?

The overflow is billed at the base you pick: each model's listed rate, its cheapest pay-as-you-go rate, or the cheaper of the two.

Can I share the comparison?

Yes. The inputs live in the URL and the Share button copies a link to the current calculation.