What does AI actually cost?
See the default estimate first, then change the text, volume or model.
Runs on worked examples — how this works
GPT-5.6 Luna
1,800 words in · 350 words out · 100 runs a day
- Per task
- US$0.00104
- Per month
- US$3.12
- Per year
- US$37
Same job. 6 configured models.
The priciest is 22.8× the cheapest for this task.
Model choice changes the operating cost
The calculator uses the same current model roster and dated prices as Model Switch. It estimates inference only, not the full cost of people, hosting, retrieval, storage or operations.
Read Reading an AI billFor the curious
Words are converted to tokens using a simple teaching estimate: one token is roughly three-quarters of a word. Real tokenisation varies by model, language and content, so production estimates should use measured provider usage.
How worked examples and data handling work
How this was built
Approach: a client-side calculator reads the shared model roster and prices. No API request is made when inputs change.
Runtime model: none. Inference cost: £0; hosting is separate.
Limit: figures are directional estimates in US dollars, not invoices or price guarantees. Arun will add the build story after review.
Build diary awaiting Arun’s notes. Empty fields below are intentional; no build story or time estimate has been invented.
- Tools used
- OpenAI Codex, Next.js 16, TypeScript, Impeccable
- Models used to build
- Arun to add
- Rough hours
- Not recorded
- What failed
- Arun to add
Actual build prompts
Yes build them into the site