Model picker

What it is

Model picker tells you the smallest kind of AI model that can do a task well: a small fast one, a mid-size one or the largest. Work that a small model handles then stops running on the dearest one out of habit.

Who it suits

People who choose models and pay the bill, and any team that runs the same AI job many times.

If you use one chat model for everything and never see a bill, skip it.

Where to use it

In any AI chat, before you choose a model. It matters most for jobs that repeat, where the gap in price between sizes adds up.

• Any AI chat: where you paste the skill’s prompt and describe the task in place of {{input}}
• A new agent or automation: ask before you set the model it runs on
• A script that calls an API for every row: the price of each call repeats there

How to set it up

What you need first:

• Where it runs: any AI chat
• A key: none
• An install: none

1. Sign in to Mindket and press Get it on this listing.
2. In your Mindket account, View shows the prompt to read and copy. This page shows it too.
3. Keep it where you can paste it again: a saved prompt in your chat app, if it keeps them, or a note.

How to use it

Paste the skill’s prompt and describe the task in a sentence or two where it says {{input}}.

1. You describe a task in place of {{input}}
2. It weighs a small, a mid-size and the largest model
3. You get one size and a line on why
4. Plus what would push the task up a size

Example

A sorting job a support team runs every day.

sort 2,000 support emails a day into billing, bug or feature request

The answer comes in three parts:

For newcomers and experts

If you are new. AI models come in sizes, and bigger ones cost more per answer and run slower. Many tasks, like sorting or tagging, do not need the biggest. Describe one job you hand to AI today and see which size it suggests.

For experts. It sorts a task into three tiers, not named models, and says what would push it up a tier. That keeps the advice stable as vendors rename and reprice models. It cannot know your latency target or your eval scores. Treat its answer as a starting point and confirm it on a sample of real inputs.

Good to know

• Sizes, not names: it names a size, not a vendor or a price, because model names and prices change often
• Your words only: it judges from your description, so a task that hides long inputs or a strict format may need more than it says
• Before you switch: try the suggested size on a handful of real inputs before you move a whole job over