Skip to content
WunderUI

Model Select

Pro

A card for choosing the model and how hard it should think: a list of models and, for the chosen one, its description, context size and price.

Playground

Claude Sonnet 5.5

The everyday model for writing, code and analysis.

Context window
1M tokens
Input
$3 per 1M tokens
Output
$15 per 1M tokens
Speed
Balanced

Reasoning effort

Balanced — the default.

Installation

Model Select comes with WunderUI Core and Pro, in the package from your plan. See the plans.

Then import what you need:

tsx
import { ModelSelect, ModelCard } from "@wunderui/react"

Then use it:

tsx
import { useState } from "react"
import { ModelSelect } from "@wunderui/react"

const models = [
  { id: "claude-opus-5-5", name: "Claude Opus 5.5", meta: "Deepest reasoning" },
  { id: "claude-sonnet-5-5", name: "Claude Sonnet 5.5", meta: "Balanced" },
  { id: "claude-haiku-5-5", name: "Claude Haiku 5.5", meta: "Fastest" },
]

export function Settings() {
  const [model, setModel] = useState("claude-sonnet-5-5")

  return <ModelSelect models={models} value={model} onValueChange={setModel} />
}

Expected result: You see a card with three models, Sonnet chosen, its name Claude Sonnet 5.5, a Reasoning effort control on Medium and the line Balanced (the default). Click Haiku and the details follow.

When to use

Use Model Select for

  • A settings page or dialog where a person chooses the model and how hard it should think
  • A choice that needs facts next to it: context window, price per 1M tokens, speed
  • A short list of models, three to eight with search, each with a description

Use something else for

  • Model Picker for the model chip of a prompt box
  • Combobox for a long list of options that is not a short model list, with type-ahead
  • Select for a plain select in a form, with a name and a placeholder
  • Usage Meter for what a run has cost so far

Details

The list shows each model's name and meta, and the pane beside it shows the description, spec rows and the reasoning effort. Leave out what you cannot say and the pane shrinks.

Claude Sonnet 5.5

The everyday model for writing, code and analysis.

Context window
1M tokens
Input
$3 per 1M tokens
Output
$15 per 1M tokens
Speed
Balanced

Reasoning effort

Balanced — the default.

Models

The card grows with the list and names wrap, so nothing scrolls or is cut off. Past five or six models, make it searchable: a field filters the list as you type.

Claude Sonnet 5.5

The everyday model for writing, code and analysis.

Context window
1M tokens
Input
$3 per 1M tokens
Output
$15 per 1M tokens
Speed
Balanced

Reasoning effort

Balanced — the default.

Control

Uncontrolled, the card keeps the model and effort itself. Controlled, your values win: if you do not change them, the card stays put, which is how an app refuses a choice.

Claude Sonnet 5.5

The everyday model for writing, code and analysis.

Context window
1M tokens
Input
$3 per 1M tokens
Output
$15 per 1M tokens
Speed
Balanced

Reasoning effort

Thinks longer on hard problems and costs more.

The card keeps the choice itself. Your code never hears of it.

Marks, bare and Model Card

A model's icon shows as a tile in the list. bare drops the frame for use inside a popover that has its own. ModelCard shows one model as a card with specs, a badge and an on/off switch.

Marks, bare in a popover, and the model card

Nova

Code and design work

Context window
200k tokens

Reasoning effort

Balanced — the default.

Keyboard interactions

The rows are buttons in a listbox with one tab stop, and the effort is a segmented control.

TaborShiftTab
Moves into the list on the chosen row (or the one you last moved to), then to the effort control, on its chosen option. The two other options are skipped.
ArrowDownorArrowUp
In the list: moves focus to the next or previous row that can be chosen. It stops at the ends. On the first row of a list, Up goes back to the search field; in the search field, Down goes to the chosen row.
HomeorEnd
In the list: moves focus to the first or last row that can be chosen.
EnterorSpace
On a row: chooses that model at once and calls . On an effort option: chooses it.
→or←
In the effort control: moves to the next or previous option and chooses it, so is called on every press. After High comes Low.
HomeorEnd
In the effort control: moves to Low or High and chooses it.

Arrow keys in the list only move focus, skipping disabled rows. A model is chosen with Enter, Space or a click, and the effort options are tabs that choose as you move.

Behavior and motion

What the card does with the model, the effort and the list, and what it leaves to you.

Not a select
ModelSelect is two panes, not a native select, with no name, form value, placeholder or groups. If you need those, build around the card or use Combobox or Select.
One model is always chosen
The card shows the value, or its own choice starting at the default or the first enabled model, and an unknown value falls back to the first. With no models the card is empty, so do not render it.
The effort
Effort is separate from the model and keeps its value when the model changes. It has three fixed values (low, medium by default, high) and three fixed hints.
What each field shows
Name and meta show in the list, with a reason under a disabled row. Name, description and specs show in the details, and icon and tag are used by Model Picker only.
Choosing
A click on a row or option chooses it at once; there is no confirmation or save. When controlled, your value wins: if your state does not change, the card stays where it was.
Search
Search adds a field above the list that filters by name, meta, description and tag. A model filtered out stays chosen with its details visible; use it from about six models.
Focus
The list is one tab stop: the chosen row, or the row you last reached with the arrow keys. Focus does not choose; Enter or Space does.
A model or an effort that cannot be used
A disabled model stays in the list, dimmed and skipped by arrow keys, with its reason under it. Effort has no disabled option: keep the old effort and show the reason next to the card.
Layout
Side by side from a 640 px viewport, stacked below it. The card grows with the list, names wrap and nothing is cut off.
The shadow
The card brings its own border and shadow. Pass a no-shadow class when it sits inside a box that already has a border.
Motion
Rows and effort options change colour quickly. Nothing else on the card moves; the effort pill does not slide.
Reduced motion
With reduced motion the colour changes keep running; there is no transform to remove.

Anatomy

The figure shows a card with specs and the Sonnet model chosen. Without a description or specs the details have fewer lines and no number for them.

Claude Sonnet 5.5

The everyday model for writing, code and analysis.

Context window
1M tokens
Input
$3 per 1M tokens
Output
$15 per 1M tokens
Speed
Balanced

Reasoning effort

Balanced — the default.

  • data-slot="model-select"

    A card with a 1 px border, 12 px radius and a shadow, up to 560 px wide. List and details sit side by side from a 640 px viewport and stack below it.

  • A listbox named Models, 208 px wide beside the details, with a divider towards them. It does not scroll, so the card grows with the list, and only one row is a tab stop.

  • Optional

    With search on: a 32 px field named Search models above the rows. A search without a match shows No model matches.

  • One model as an option button: the name in 14 px medium and the meta below in 12 px. The chosen row has the subtle fill, and a disabled row is dimmed with its reason under it.

  • The name of the chosen model, 16 px semibold, at the top of the details.

  • Optional

    The description of the chosen model in 12 px, if it has one.

  • Optional

    The specs as rows of a description list: label on the left, value on the right in medium weight. Not drawn without specs.

  • The Reasoning effort label and a small segmented control named the same, with Low, Medium and High. The options are announced as tabs.

  • One 12 px line for the chosen effort: Quick answers, lowest cost; Balanced (the default); or Thinks longer on hard problems and costs more. The words are fixed.

Best practices

  • Do

    Give each model a short meta for the list and put compared facts into specs: context window, price per 1M tokens, speed. Use the same labels for every model so values line up.

  • Do

    Keep the choice in your state and say what it changes, such as Applies to new chats. The card is not a form control: no name, no form value, no submit.

  • Do

    Disable a model that cannot be used, give a reason and keep the old value. For an unusable effort, say so next to the card and offer a way out: another model, or the plan that includes it.

  • Do

    Order the list the way people choose: most capable or recommended first, not alphabetically.

  • Don’t

    Don't use it for a long list without search. Rows do not scroll, so the card grows with every model; past about eight models, or when you need groups, use Combobox.

  • Don’t

    Don't rely on effort for a model that does not think longer. Effort is separate from the model and never changes when it does; reset it yourself if a model needs it.

  • Don’t

    Don't start a request when the model or effort changes. Picking is immediate with no confirmation; save when the person says so.

  • Don’t

    Don't disable a model without a reason. A dimmed row with no explanation looks like a bug; say what to do, such as Rate limit reached · back in a minute.

API reference

ModelSelect is a plain component, so this table lists every prop. A ModelOption has an id and a name, plus optional meta, description, tag, icon, specs, disabled and reason.

A card with a 1 px border, 12 px radius and a shadow, up to 560 px wide. List and details sit side by side from a 640 px viewport and stack below it.

PropTypeDefaultDescription
models requiredModelOption[]—The models, in the order of the list: . is the value. shows under the name in the list; and show in the details. keeps the row visible but not choosable and says why. and belong to Model Picker and are ignored here.
valuestring—The id of the chosen model, for a controlled card. If you do not change it in , the card stays where it was.
defaultValuestring—The id that starts chosen when you do not pass . The first model by default.
onValueChange(id) => void—Called with the id of the row that was clicked, also for the row that is already chosen.
effort"medium" | "low" | "high"—, or , for a controlled card. If you do not change it in , the effort stays where it was.
defaultEffort"medium" | "low" | "high""medium"The effort that starts chosen when you do not pass . by default.
onEffortChange(effort) => void—Called with the new effort on every click and every arrow key in the effort control.
searchablebooleanfalseAdds a search field above the list that filters by name, meta, description and tag. Default false.
barebooleanfalseDrops the border, radius and shadow of the card, for use inside a popover or sheet that has its own frame (no double frame).

PropTypeDefaultDescription
model requiredModelOption—
badgeReactNode—Replaces the badge. Default: "Off" while turned off, else .
enabledbooleantrueWhether the model is on. Default true.
enabledLabelstring"On"Text beside the switch. Default "On".
footerReactNode—Small print at the start of the footer, e.g. "1,204 requests".
onEnabledChange(enabled) => void—Shows the on/off switch in the footer.
onSelect() => void—Makes the name a button that covers the card.
selectedbooleanfalseThe model that new chats start with: primary rim and subtle background.
specs[string, ReactNode][]—Label and value pairs in two columns. Default: .

Notes for AI

Give your coding assistant the Markdown reference instead of screenshots.

MCPget_component("model-select")
InstallIncluded in Core and Pro

The full library index for assistants is at /llms.txt.