Models and your allowance

Two things decide how Lucy feels day to day: which model answers you and how much you can use. Both are designed so you rarely have to think about them, and can take control when you need to.

Which model answers

Auto: the default

In Auto, Lucy chooses the model for each answer. A quick question gets a fast, light model. A long contract or a careful analysis gets a stronger one. You get good answers without learning model names, and your organisation does not pay for a large model on every small request.

Choosing a model yourself

If your organisation allows it, the model picker appears in the box where you write your message.

• Auto (recommended) lets Lucy decide for each answer.
• A named model is used for every later answer in this conversation. The picker says “Applies from the next answer”.
• The dots next to a model show how fast it uses your allowance: one dot is light, three dots use it fastest. They are relative, never amounts of money.
• Ask your admins lists models you have not been given. Choose one and send a short reason; your admin approves or declines.

Which model for which job

Job	A good choice
Short questions, quick drafts, rewrites	A small, fast model, or Auto
Everyday documents, summaries, spreadsheets	A medium model, or Auto
Long contracts, many files, careful comparison	The strongest model your organisation offers
Charts, scans and screenshots	A model that reads images (the picker says so)

When your choice cannot be used

If the model you chose cannot do something, Lucy does not fail. It picks another model for that answer and tells you why:

• the conversation holds an image and the model cannot read images;
• the conversation has become longer than the model can hold.

Your choice stays in place for later answers where it works.

Your allowance

Your organisation can give each person an allowance, so costs stay predictable and nobody is surprised by a bill. It works in three windows at once.

Window	What it means	Resets
Current session	A rolling five hours that starts with your first answer. It stops one long burst from using everything.	Hours after your first answer, shown on the page
This week	Your share for the week	Monday, shown on the page
This month	Usually your organisation’s shared monthly budget	The 1st of the month

You see percentages and reset times, never amounts of money. When any bar is full, new answers wait until it resets or until you get more.

What uses your allowance faster

• Longer answers and longer conversations, because Lucy rereads more.
• Stronger models, which is what the dots in the picker show.
• Work in the sandbox, such as analysing a big spreadsheet or building a deck.
• Web searches, if your organisation enabled them.

Short, specific requests in a fresh conversation go furthest. Start a new conversation when you change topic.

Your last 30 days

Below the bars the page shows your own activity: answers, sandbox time, web searches and stored data, per day. Use it to see what your work actually costs and to spot a conversation that has grown too long.

Need more?

1. On the Usage page, choose Ask for more next to the window that is full.
2. Write a short reason, for example “month-end close, reconciling 14 accounts”.
3. Your admin sees the request, approves or declines, and leaves a note. If approved, your new limit applies straight away.

The same flow works for a model you have not been given, from the Ask your admins section of the picker.

Making your allowance last

Put the whole job in one clear request instead of many small ones. Attach the files once. Pick a stronger model only for the step that needs it, then switch back to Auto.

Admins: see Limits and budgets and Models for how these are set.