OpenTahr

Home Blog AI without a subscription

AI without a subscription

Use AI with your own API keys, from getting the keys to real costs and the right settings.

A woman printing an orange pattern with a wooden block in her studio

Prompt caching cuts what you pay for repeated input

Prompt caching bills your repeated input at a fraction of the price. The exact conditions per provider, the golden writing rule and how to check the savings.

Read the article →
A man turning a dial on a wall of large gauges in a workshop

OpenAI spending limit with project budgets and alerts

Cap your OpenAI spending in one minute with an organization limit, a budget per project and alert thresholds, then watch out for automatic recharge.

Read the article →
A woman sorting letters into the pigeonholes of a wooden mail cabinet

AI model routing, send simple tasks to cheaper models

Send each task to the model tier that is enough, with fixed rules, a cascade or a classifier. A task to model table, the practical setup and the traps to avoid.

Read the article →
A gardener pruning a small potted olive tree with shears

How to reduce AI costs with 8 levers ranked by gain

Eight levers reduce AI costs without losing quality, ranked by gain against effort, from picking the model to reading the statement, each with its detailed guide.

Read the article →
A woman sorting coins and notes into labeled glass jars on a kitchen table

Monthly AI budget, how to estimate it accurately

A monthly AI budget comes down to one formula and four steps. Three budgets worked out to the cent and a spending cap turn the estimate into a guarantee.

Read the article →
A smiling woman picking a pen out of a tray at a flea market stall

The cheapest AI model for writing, ranked by price

Claude Haiku 5.5 and GPT-6 Luna lead the cheapest AI models for writing at $0.50 per million output tokens, with a dated ranking and a typical article costed on each.

Read the article →
A cheesemonger holding up two pieces of cheese at his market stall

The best value AI model for each kind of user

Value in AI depends on who you are. Mid-tier models suit daily work, floor models suit volume and flagships suit brand pages, with dated verdicts and a method.

Read the article →
A woman comparing price labels at a fruit and vegetable market stall

AI API pricing compared on one dated table

AI API pricing for 22 models from 6 providers, all recorded on a single day. The full tables come with recalculated simulations and the hidden costs that skew comparisons.

Read the article →
A man reading a printed page while taking notes at his desk

Input vs output tokens and why AI output costs more

AI output costs two to six times the input because it is generated token by token. The explanation comes with ratios by provider and four useful habits.

Read the article →
A woman reading a recipe card while weighing flour on a kitchen scale

Calculate the API cost of an AI call, step by step

The API cost of an AI call comes from a one-line formula. Token counting, dated official prices and a full worked example take it all the way to a monthly budget.

Read the article →
Hands slicing a baguette into even pieces on a wooden board

What an AI token is and what it really costs

A token is a word fragment and the unit AI models read and bill in. Concrete equivalents, the case of French and dated official price lists put it in perspective.

Read the article →
A smiling woman carrying a cardboard box and a potted plant through a doorway

Switch from an AI subscription to an API key, step by step

Moving from an AI subscription to your own API key means cancelling cleanly, creating the key and rebuilding your uses. A seven-step checklist covers it, along with the cases where you should stay.

Read the article →