> ## Documentation Index > Fetch the complete documentation index at: https://docs.bezi.com/llms.txt > Use this file to discover all available pages before exploring further. # Model Selection You can pick the model Bezi uses to affects the depth, speed, and cost of a response. How to decide: **heavier tasks need more reasoning power; simple tasks don't**. The right model for the right task gets you a good result faster, which will help credits go further. ## Model type options Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: 1. **Featured Models**: our team has confirmed these are reliable and support all the functionality Bezi needs to perform any game development related workflow. 2. **Experimental Models**: riskier and generally less expensive. Some may not support image attachments or other features Bezi needs in order to execute tasks to satisfaction. ## When to use each model type A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: 1. **Featured Models**: our team has confirmed these are reliable and support all the functionality Bezi needs to perform any game development related workflow. 2. **Experimental Models**: riskier and generally less expensive. Some may not support image attachments or other features Bezi needs in order to execute tasks to satisfaction. ## When to use each model type A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: 1. **Featured Models**: our team has confirmed these are reliable and support all the functionality Bezi needs to perform any game development related workflow. 2. **Experimental Models**: riskier and generally less expensive. Some may not support image attachments or other features Bezi needs in order to execute tasks to satisfaction. ## When to use each model type A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. Bezi offers a constantly-updating selection of models - from heavy, frontier models to lightweight, experimental ones - so you can pick the one that works best for you. All models are organized into one of these categories: 1. **Featured Models**: our team has confirmed these are reliable and support all the functionality Bezi needs to perform any game development related workflow. 2. **Experimental Models**: riskier and generally less expensive. Some may not support image attachments or other features Bezi needs in order to execute tasks to satisfaction. ## When to use each model type A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. A good rule of thumb: **start with a featured model**. Test experimental models once you're familiar with Bezi's platform, AI behavior and quirks and have a new workflow you're comfortable in. These are more expensive but well worth the extra credits for time saved on complex tasks: * Long, multi-step Agent Mode workflows * Output that requires precision, like UI generation * Tasks with many dependencies or subtle system interactions * Deep debugging across scenes and systems where the root isn’t clear Genuinely great for most daily work: * Debugging and error fixing * Writing or modifying scripts * Building an editor tool Best for getting a fast answer or iterating on lighter tasks: * Quick lookups and explanations * Simple one-line fixes you're confident about * Tasks where you plan to review and refine the output anyway ## How to change the model selection The model selector is inline in the prompt input. To select a different model type, open the dropdown and pick the new model before you submit a prompt. ## Tracking credit usage To see a breakdown of credits consumed per prompt and model type, follow these steps: 1. Go to: [app.bezi.com/account](https://app.bezi.com/account) 2. Open the **Subscriptions & Usage** tab See the [Plans & Credits](https://docs.bezi.com/account/plans) page for a full breakdown of how credit usage works.