> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hiveku.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Models

> Mini, Max, and Ultra tiers

Hiveku Builder offers three AI model tiers. Each tier balances speed, capability, and cost differently so you can pick the right one for the task at hand.

## Model tiers

| Tier      | Availability               | Best for                                                                                                  |
| --------- | -------------------------- | --------------------------------------------------------------------------------------------------------- |
| **Mini**  | All plans (free)           | Quick edits, simple questions, single-file changes                                                        |
| **Max**   | Standard and Premium plans | Feature development, debugging, multi-file changes, day-to-day coding work                                |
| **Ultra** | Premium plan only          | Complex architecture decisions, large multi-file refactors, deep parallel exploration, advanced reasoning |

## When to use each tier

**Mini** is the right default for:

* Renaming variables, fixing a typo, tweaking copy
* Quick "what does this function do?" questions
* Small CSS / styling adjustments

**Max** is the right default for:

* Building a new feature across a few files
* Debugging an issue with specific reproduction steps
* Adding a new page, component, API route
* Refactors that touch 2-5 files
* Most everyday coding work

**Ultra** is worth reaching for when:

* A refactor will touch many files and the reasoning needs to stay coherent across the entire change
* Architecture decisions where the tradeoffs need careful weighing across multiple options
* Deep debugging where the cause may be several layers removed from the symptom
* Large parallel exploration ("find every place we use X, understand the pattern, propose a unified replacement")

A useful test: if your task feels like *"this is going to take an hour of my undivided focus to think through carefully"*, it's probably an Ultra task. If it feels like *"I know what to do, I just need help executing it"*, Max is plenty.

## Reasoning effort

Each model tier supports three reasoning effort levels:

* **Low** — minimal thinking, fastest responses
* **Medium** — balanced depth and speed (default)
* **High** — deep analysis, slower but more thorough

Higher reasoning effort consumes more credits per message on the Max and Ultra tiers.

## Switching models

Select your model tier from the dropdown in the chat panel toolbar. The change takes effect on the next message you send.

<Info>
  You can switch models mid-conversation. The AI retains the full context of your chat regardless of which model you use. A common pattern: explore and design on **Ultra**, then switch to **Max** to execute the plan once direction is clear.
</Info>

## Credit consumption

* **Mini** — free on all plans, no credit deduction
* **Max** — deducts credits based on message complexity and reasoning effort
* **Ultra** — deducts credits at a higher rate, reflecting the more powerful model

See [AI Credits](/settings/ai-credits) for a full breakdown of credit consumption and plan allowances.