TUE, NOVEMBER 10, 2026
Independent · In‑Depth · Practitioner‑Tested
Claude LLMs

Budget Model Prompts: Getting Real Work Out of $0.10 Models

Muse Ember, Claude Haiku 5.5 and gpt-6-luna all cost $0.10 per million input tokens. At that price you can afford to run them on everything - but only if the prompt does the thinking the model will not. These eight are written for cheap models doing volume: narrow scope, explicit output shape, no open-ended reasoning.

⌨️ 8 prompts 🕐 Updated Nov 10, 2026 👁 5 views
💡 How to use these prompts: Replace everything in [BRACKETS] with your specific details before sending. Click Copy to copy any prompt to your clipboard instantly.
1
Classify into a fixed set
Support triage, feedback routing
Classify the text below into exactly one of these categories: BUG, FEATURE_REQUEST, QUESTION, PRAISE, COMPLAINT.\n\nReturn only the category name. No explanation. If the text fits none of them, return OTHER.\n\nText:\n[PASTE]
2
Extract structured fields
Data entry, document processing
Extract these fields from the text below and return them as JSON with exactly these keys: company, product, price_usd, date_iso.\n\nIf a field is not present, use null. Do not infer or estimate. Return only the JSON object.\n\nText:\n[PASTE]
3
Rewrite to a fixed length
Excerpts, meta descriptions, previews
Rewrite the text below in exactly two sentences, maximum 40 words total.\n\nKeep every number and proper noun unchanged. Do not add information that is not in the original. Return only the rewritten text.\n\nText:\n[PASTE]
4
Yes or no with a reason
Compliance checks, content gating
Answer the question below with YES or NO on the first line, then one sentence of reason on the second line. Nothing else.\n\nIf the text does not contain enough information to decide, answer UNKNOWN.\n\nQuestion: [QUESTION]\nText:\n[PASTE]
5
Normalise messy input
CRM cleanup, list hygiene
Convert the list below into clean rows. Each row: Name | Email | Company.\n\nFix obvious capitalisation. Do not invent missing values, leave them blank. Do not reorder. Return only the rows, one per line.\n\nList:\n[PASTE]
6
Score against one criterion
Editorial QA, bulk content review
Score the text below from 1 to 5 on how clearly it explains its main point, where 1 is unclear and 5 is immediately clear.\n\nReturn only the number and then, on a new line, the single sentence that most needs rewriting.\n\nText:\n[PASTE]
7
Tag with a controlled vocabulary
Content tagging, review analysis
Tag the text below using only tags from this list: pricing, security, performance, usability, support, documentation.\n\nReturn between one and three tags, comma separated, most relevant first. Use no other tags. Return only the tags.\n\nText:\n[PASTE]
8
Detect a single condition
Pricing extraction, document scanning
Read the text below and answer one question: does it contain a specific dollar amount?\n\nIf yes, return the amount exactly as written. If no, return NONE. Return nothing else.\n\nText:\n[PASTE]