$20 Classroom AI Budget: A $14.40 Planning Receipt

2026-10-26 Spend less EN

01 / A $20 ceiling for 30 students and four lessons

A classroom AI experiment needs a spending limit before it needs a clever prompt. For this plan, the ceiling is $20 for 30 students across four lessons. Each student gets six AI exchanges per lesson: enough to request feedback, challenge it, and revise without turning the activity into unlimited chat.

The receipt below totals $14.40, including a retry allowance. That leaves $5.60 unallocated. These are planning figures, not a claim about CheapRelay’s current prices. The example uses assumed rates of $5 per million input tokens and $15 per million output tokens. Replace both with the selected model’s current rates before approving the activity.

02 / The student-use receipt comes to $10.80

One exchange means one API request and its response. Budget an average of 1,500 input tokens and 500 output tokens per request. Input includes instructions, the student’s text, and any conversation history resent with the request—not merely the newest question.

■ Students: 30.
■ Lessons per student: 4.
■ Exchanges per lesson: 6.
■ Total student requests: 30 × 4 × 6 = 720.
■ Input tokens: 720 × 1,500 = 1,080,000.
■ Output tokens: 720 × 500 = 360,000.
■ Input charge: 1.08 million × $5 = $5.40.
■ Output charge: 0.36 million × $15 = $5.40.
■ Student-use subtotal: $10.80.

This is a token estimate, not a promise that every response will have the same length. A long pasted essay or repeated conversation history can push input usage above the allowance. Set an output-token limit, but remember that a response limit alone does not control input costs.

03 / Teacher preparation and retries bring the plan to $14.40

Reserve 40 additional requests for testing instructions, checking feedback quality, and preparing examples. Allow 3,000 input tokens and 1,000 output tokens per preparation request.

■ Teacher input: 40 × 3,000 = 120,000 tokens; estimated charge $0.60.
■ Teacher output: 40 × 1,000 = 40,000 tokens; estimated charge $0.60.
■ Teacher preparation subtotal: $1.20.
■ Student use plus preparation: $12.00.
■ Retry and revision allowance: 20% × $12.00 = $2.40.
■ Planned total: $14.40.
■ Unallocated portion of the $20 ceiling: $5.60.

The retry allowance covers extra usage, not a guaranteed number of failed requests. Likewise, the remaining $5.60 is a buffer, not permission to extend every conversation. If actual rates or token counts are higher, reduce the activity before students begin.

04 / Keep the API key behind a classroom gateway

CheapRelay offers pay-as-you-go access through one key, with no monthly fee or subscription. Its OpenAI-compatible routes sit under /v1 at https://api.cheapapirelay.com. That can simplify a small teacher-managed application, but compatibility does not replace spending controls.

Keep the key on a server, never in student-facing browser code. Give each student a classroom login and enforce six exchanges per lesson. Limit pasted text, shorten retained history, and reserve the maximum estimated charge before accepting each request. Reconcile that reservation against actual usage afterward, using current billing rates rather than this article’s sample rates.

CheapRelay bills real tokens, including cache reads and writes. This receipt assumes ordinary input and output usage without separate cache charges or savings. Check the selected model’s billing details and add any applicable categories; do not silently count caching as free.

05 / Spend the six exchanges on revision, not answer collection

Ask students to draft first, request criticism second, and submit a revision with an explanation of which suggestions they accepted. Use anonymized practice material rather than names, grades, or sensitive student records. Follow your institution’s approval and consent requirements before sending classroom work to an external service.

Run a small teacher-only pilot, inspect token usage, then recalculate the receipt. Sign up for CheapRelay when you are ready to test that pilot against current prices. The useful outcome is not maximum token consumption: it is a completed learning task with a bill you can explain line by line.

More reading

Date	Aisle	Written in	Title
2026-10-26	Spend less	EN	$30 a Month for AI: An Indie Hacker’s Receipt A $30 monthly AI budget with token math, cache allowances, retry limits, and practical spending controls for solo builders.
2026-10-26	Spend less	EN	$3.45 for 95 accepted answers beats cheap tokens A worked receipt shows why token prices can mislead: count retries, rejected outputs and cache charges before ranking models.
2026-10-26	Spend less	EN	The $0.004 AI Task With a $0.0566 Agent Receipt A hypothetical agent receipt shows how context, retries, evaluation and logging turn a $0.004 task into $0.0566 per accepted result.
2026-10-26	Spend less	EN	$2.53 for 10,000 JSON Jobs: Read the Receipt A worked JSON-output budget: token costs, schema overhead, retries, and the checks that keep cheap API calls from becoming expensive workflows.
2026-10-26	Spend less	EN	$0.012 Either Way: Streaming Changes the Habit A token-by-token receipt shows why streaming has no automatic discount, but stopping early and avoiding retries can change your API bill.
2026-10-26	Spend less	EN	$1.71 to $1.02: A Long-Chat Summarisation Receipt A worked token receipt shows when summarising chat history saves money, including the summary fee, break-even point and caching caveats.