AI models advertise low per-token prices, but real projects burn tokens on prompts, retries, and long outputs. CostPerTask is a free browser estimator that helps you compare what a recurring AI task might cost across models — using prices you type from official vendor pages.
Open the tool: https://mxhoaib.github.io/costpertask/. This tutorial shows how to use it safely, where to find official pricing, and how freelancers and project managers can turn rough token maths into a monthly budget line.
What CostPerTask is (and is not)
CostPerTask runs in your browser. You pick a task profile (input tokens, output tokens, tasks per month), then edit model rows with USD prices per 1 million input and output tokens. The table recalculates per-task and monthly cost. You can add models, reset examples, copy results, or download a CSV.
- It is: an editable estimator for planning and teaching.
- It is not: a quote, invoice, or financial advice.
- Prices: example placeholders until you overwrite them from official pricing pages.
- Privacy: nothing is stored or sent for the estimator itself.
Token counts in the examples are rough (about 1 token ≈ ¾ of an English word). Always validate against your own logs when you have them.
How to estimate a task in five steps
- Visit CostPerTask.
- Set input tokens per task, output tokens per task, and tasks per month for a realistic workflow (for example: weekly proposal drafting, daily RFI summaries, or batch captioning).
- Open each model’s official pricing page from the tool’s links (or search the vendor site). Enter current USD prices per 1M input and 1M output tokens.
- Compare per-task and monthly columns. Add or remove models as needed.
- Copy results or download CSV for a client rate card, PMP cost baseline, or team Slack note.
Recommended official pricing sources (verify dates yourself — prices change):
Always prefer the vendor’s own documentation over third-party mirrors or “free key” blogs. That keeps your research AdSense-safe and legally clean.
Why freelancers and PMP-minded teams need this
Drafting studios, content freelancers, and UAE consultants increasingly mix AI into proposals and documentation. CostPerTask helps you:
- Stop guessing whether a “cheap” model stays cheap at high monthly volume.
- Explain AI line items to clients without inventing numbers.
- Compare models when output length (not just input) drives the bill.
- Build a simple cost baseline for project planning documents.
If you produce AutoCAD deliverables or marble shop drawings professionally, treat AI chat cost as a separate overhead from drafting hours — then decide where human CAD time still creates the value clients pay for.
Practical example workflow
Suppose you summarise ten site meeting notes per month. You estimate 2,000 input tokens and 800 output tokens per summary. Plug those numbers into CostPerTask, paste current official rates for two or three models, and you instantly see which option is cheaper at that volume. Raise tasks-per-month to 100 and watch the ranking change. That sensitivity check is the real teaching point of the tool.
Optional paid help listed on the tool page covers a custom cost report for exact workflows. For drafting production itself, contact D3DDD rather than expecting an estimator to replace CAD.
Responsible use checklist
- Update prices from official vendor pages before you share numbers with a client.
- Disclose that figures are estimates based on assumed token counts.
- Do not paste confidential drawings or personal data into public AI chats if contracts forbid it.
- Never use cracked software or leaked API keys to “save cost.”
Building a reusable cost template
Once you trust your token assumptions, save a CSV from CostPerTask and keep a short “assumptions” note beside it: date checked, pricing URLs used, and whether taxes or seat fees are excluded. Model API prices are only one part of total cost of ownership. You may also pay for IDE copilots, image generators, or storage. Keep those on a separate spreadsheet tab so the per-token estimator stays clean.
For agencies, create one row set per workflow: proposal writing, meeting summary, code assist, and image captioning. Different workflows have different input/output ratios. Captioning might be output-heavy; summarising long PDFs is input-heavy. CostPerTask makes that ratio visible when you edit token fields deliberately instead of leaving placeholders unchanged.
Teaching clients without overpromising
When a client asks “can’t AI do this for free?”, show a CostPerTask screenshot with official prices filled in and a realistic monthly volume. Transparency builds trust faster than vague claims. Then separate AI overhead from professional drafting fees. Floor plans, elevations, and marble shop drawings still require human QA, layer standards, and liability awareness — especially on UAE fit-out projects with fabricators waiting on cut lists.
Project managers studying PMP-style cost management can treat the CSV as a lightweight cost-estimating artefact: basis of estimate, units (tokens), rates (USD per 1M), and quantity (tasks per month). Refresh rates monthly because vendors revise pricing.
Common mistakes to avoid
- Using blog-scraped prices instead of official vendor pages.
- Forgetting that retries and tool-calling multiply tokens.
- Comparing a tiny demo prompt to a production system prompt with long memory.
- Ignoring currency conversion and VAT when reporting to UAE clients.
- Storing API keys in public repos while “optimising cost.”
Primary CTA: Run your estimate free at https://mxhoaib.github.io/costpertask/.
Soft CTA: Need human drafting, 3D, or marble shop drawings after the planning stage? info@d3ddd.com · d3ddd.com.
About the author — Shoaib Hussan: Founder of D3DDD (Dubai 3D Design & Draft). Shares practical AutoCAD, civil drafting, marble shop drawings, and AI workflow tips for UAE projects. Instagram: @mxhaib. Contact: info@d3ddd.com · d3ddd.com.
ConversionConversion EmoticonEmoticon