> ## Documentation Index
> Fetch the complete documentation index at: https://shapesinc-4644c49f-cursor-docs-product-refresh-1067.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Which AI Model Should I Use?

> A fast guide to picking the AI model behind your Shape: by what you're doing, free vs premium, and the badges that matter. 300+ models, most free.

Every Shape runs on an AI **model** (we also call it an engine). The same character feels different depending on which one you pick. There are 300+ to choose from, and most are free.

You don't need to memorize model names. Pick one by what you're doing, try it in chat, and swap it if it feels off.

<Tip>
  This is the fast "what do I pick?" guide. Want the family-by-family tour (Claude vs GPT vs Gemini vs DeepSeek and friends)? Read the [AI Engine Guide](/aienginecheatsheet). The live list is always at [shapes.inc/engines](https://shapes.inc/engines).
</Tip>

## Pick by what you're doing

| If you want…                     | Try                                                 | Why                                                   |
| -------------------------------- | --------------------------------------------------- | ----------------------------------------------------- |
| **Writing and roleplay**         | Claude, GLM, DeepSeek, MiniMax                      | Expressive, stays in character, great at long scenes. |
| **A smart all-rounder**          | Claude Sonnet, Gemini, Kimi, GPT                    | Reliable, follows directions, handles mixed tasks.    |
| **Coding and reasoning**         | GPT, Claude, DeepSeek, Qwen, Gemini Pro             | Careful and structured, good at hard problems.        |
| **Free daily chat**              | Gemma, Gemini Flash, GPT Mini, Llama, Mistral, Qwen | Fast, capable, free.                                  |
| **Looser, more surprising chat** | Gemini, Grok, DeepSeek                              | Less filtered, more playful.                          |
| **Understanding images**         | Anything with the **Native Vision** badge           | Reads pictures people send.                           |

Treat these as starting points. The best engine for a Shape is the one that feels right when you actually chat with it.

## Free vs premium

<CardGroup cols={2}>
  <Card title="Free engines" icon="gift">
    Cost nothing. A strong free engine is always available, so you never have to pay to keep chatting.
  </Card>

  <Card title="Premium engines" icon="gem">
    The top-tier models. They cost [Shape Credits](/credits), which you can buy or earn free through [rewards](/rewards). A chat can also be [sponsored](/credits-guide) so everyone uses premium for free.
  </Card>
</CardGroup>

On the [engines page](https://shapes.inc/engines), the **Always Free** count shows how many engines cost no credits right now. Filter to **Free** to see only those.

## Read the badges

Each engine carries badges that tell you what it's good at:

* **Reasoning:** better at careful thinking, planning, coding, and hard questions.
* **Native Vision:** can understand images.
* **Tools:** can use [Shape Skills](/shapeskills) like web search and code.
* **Free:** costs no credits.
* **Premium:** costs Shape Credits.
* **New:** added recently.
* **Unstable:** might fail or act weird. Fine to experiment with. Don't make it a Shape's main engine.

## Set a primary and a fallback

In a Shape's [AI Engine settings](/shape-settings#ai-engine) you pick two:

* **Primary:** the engine that powers your Shape day to day.
* **Fallback:** what it switches to if the primary is briefly unavailable, so a conversation never stalls.

Pick a strong primary for the vibe you want, and a dependable free engine as the fallback.

## How to actually decide

<Steps>
  <Step title="Start with a default">
    Use a well-rounded free engine. It's good enough to get your Shape talking.
  </Step>

  <Step title="Match it to the job">
    Use the table above. Roleplay Shape? Try an expressive model. Coding mentor? Try a reasoning model.
  </Step>

  <Step title="Test it in a real chat">
    Send a handful of messages. Does it sound right? Too long, too flat, too filtered? [Regenerate](/regeneration) a few replies to feel its range.
  </Step>

  <Step title="Swap freely">
    Changing engines is one click and doesn't reset your Shape. Try two or three before you settle.
  </Step>
</Steps>

<CardGroup cols={2}>
  <Card title="Family-by-family guide" icon="layer-group" href="/aienginecheatsheet">
    The deeper tour of every model family and badge.
  </Card>

  <Card title="Browse the live list" icon="microchip" href="https://shapes.inc/engines">
    Every engine available right now.
  </Card>
</CardGroup>

[Pick an engine on Shapes](https://shapes.inc/engines)