Skip to main content

Choose an AI model and see what it costs

Where to pick the model for an AI column, how the price per row and the tiers are shown, and where to compare models.

Written by Umer Ishaq

An AI column runs on a model you choose. Every model in the picker shows an estimated price per row.

Where do I choose the model?

  1. Click Add column, then Use AI.

  2. Open the Configure tab.

  3. Under Model, click the model menu. A new column starts on Loopmodel OG.

Next to Model there is a Compare models link.

The Configure tab of the Use AI panel, with the Generate and Configure tabs at the top. It shows a Loopgent menu set to Custom prompt, a Use case switch with Text/Data selected, a Compare models link, a Model menu showing Loopmodel OG, an empty Prompt box, a Test prompt section and Define outputs below. The Model menu is outlined in red.

What is the "You probably don't need expensive models!" window?

When you pick a model, this window can open. It says: "Cost of AI enrichment is dynamically calculated based on the model and the data." It shows cards with a cost per enrichment, for example for Loopmodel OG and GPT 5.2. It also says you are free to use any model and that this is just a heads up.

  • Don't show this again is a tick box.

  • Click Got it, continue to go on. The model you picked applies after that.

The You probably don't need expensive models window. It says the cost of AI enrichment is calculated from the model and the data, shows two cards with a cost per enrichment for Loopmodel OG and GPT 5.2, a short note that Loopmodel OG is optimized for data enrichment, an empty Don't show this again tick box and a green Got it, continue button. The Got it, continue button is outlined in red.

What does the picker show?

Each model has a name, an approximate price per row (marked with ~), a tier label and one line about what it is for. The tiers are Cheap, Efficient, Standard, Expensive and Ultra Expensive.

The prices depend on the Use case. These are the Text/Data prices for the Loopster models:

  • Loopmodel OG (Recommended): about $0.0027 per row, Standard. "Balanced default for most columns — strong quality at a low price."

  • Loopmodel Mini: about $0.00034 per row, Cheap. "Cheapest option for simple, well-defined extraction at high volume."

  • Loopmodel Max: about $0.018 per row, Ultra Expensive. "Strongest option for deep research and complex multi-step analysis."

The picker also lists models from other providers, each with its own price and tier. In Text/Data, for example, GPT 6.1 Sol shows about $0.0073 per row and GPT 5 Nano about $0.00022 per row. The list changes, so check the picker for the current models and prices. At the bottom of the menu, a line reads "How pricing works" with a link, and Compare models.

The Model menu of the Use AI panel open on the Text/Data use case. Each model row shows a name, an approximate price per row marked with a tilde, a tier label such as Cheap, Standard, Expensive or Ultra Expensive, and a one-line description. The list starts with Loopmodel OG marked Recommended, then Loopmodel Mini, Loopmodel Max and GPT models. The Loopmodel OG row is outlined in red. Part of the Use AI panel with a Compare models link is visible on the right.

What do the prices look like with Web Search?

When Use case is Web Search, the picker shows different numbers:

  • Loopmodel OG: about $0.183 per row, Standard.

  • Loopmodel Mini: about $0.024 per row.

  • Loopmodel Max: about $1.217 per row.

  • GPT 5 Nano: about $0.013 per row.

  • Claude Opus 5: about $1.217 per row.

Check the Use case before you read a price.

Where do I see the price before I run?

The footer of the Use AI panel shows no price. Look here instead:

  • In the model list, as an approximate price per row.

  • In the Run column dialog, before anything runs: an "Expected cost: ~$…" line under Run all N empty rows and under Overwrite all N rows.

The Run column dialog with a purple Run all 4 empty rows button, an Expected cost line of about $0.01 under it, and the note Pick this if you're sure your setup is correct. Below is a red Overwrite all 5 rows button with its own Expected cost line and the note Re-run every row, overwriting completed cells. The Expected cost line under Run all 4 empty rows is outlined in red.

What else is on the Configure tab?

  • Use case: Text/Data or Web Search. Text/Data says "Loopgent runs strictly on provided input cells, without internet access." Web Search says "Loopgent researches each row — Google search and webpage scraping."

  • Prompt: required. Type @ or / to tag a column.

  • Test prompt: runs the Loopgent on the first row, "without saving results or spending from your wallet". The result shows in the panel as a card: Row 1, Complete, the answer and a collapsible Loopgent trace. On a Text/Data column, from Edit column, nothing is billed: the wallet and Usage history do not change, and the table's cells stay empty.

  • Define outputs: AI Response, Output format (Fields or JSON Schema) and + Add output.

Good to know

Related Articles
Did this answer your question?