> ## Documentation Index > Fetch the complete documentation index at: https://docs.replit.com/llms.txt > Use this file to discover all available pages before exploring further. # Model selector > Use intelligent model routing or choose an available primary model for Replit Agent. The Model selector lets Replit use intelligent model routing or lets you select an available model yourself. ## Intelligent model routing Intelligent model routing chooses an appropriate model for each task, balancing quality and cost. When you see **Auto** in the product, Replit is performing intelligent model routing. Intelligent model routing considers the work you request and its complexity. A focused request can use a lower-cost model, while complex work may need a more capable model. Replit can select a different model as the task changes. Free Mode always uses intelligent model routing. Builders in Free Mode cannot use the Model selector or choose a model manually. ## Choose a model manually Builders on Core and Pro can choose an available model in Power Mode or Max Mode. 1. In the prompt box, select **Agent Mode**. The menu shows the available modes: **Free**, **Power**, and **Max**. 2. Select **Power** or **Max**, then find the **Primary model** section. 3. Open the primary model menu, then choose a model from the list. Agent modes popover showing Free, Power, and Max modes with Power selected, a Primary model list including GPT-5.6 Luna Fast and Claude Sonnet models, and an Effort slider set to High Agent modes popover showing Free, Power, and Max modes with Max selected, a Primary model list including Claude Fable 5, Claude Opus 5, GPT-5.6 Sol, and Kimi K3, and an Effort slider set to Low Your selection applies to your next message; you can change it at any time. ## Models by mode The Model selector offers the following models in Power Mode and Max Mode: | Model | Power Mode | Max Mode | | - | :-: | :-: | | GPT-6 Sol | ✓ | | | GPT-6 Luna Fast | ✓ | | | Claude Sonnet 5 | ✓ | | | Claude Sonnet 4.6 | ✓ | | | GPT-6 Astra | | ✓ | | Claude Fable 5.1 | | ✓ | | Claude Opus 5.5 | | ✓ | | Claude Opus 4.8 | | ✓ | | Claude Opus 5 | | ✓ | | Claude Opus 5 Fast | | ✓ | | Kimi K3 | | ✓ | Free Mode does not have a selectable model list. The list in your Model selector is the source of truth for your account. Model selection is rolling out, and the available models can vary by rollout, organization settings, and authorization. The **Fast** model variants require a Pro or Enterprise plan. If you don't see **Primary model**, the Model selector isn't available for your account yet. ## Effort **Effort is a per-model control, not a separate mode.** It sets how much reasoning the selected model spends on each response, from **Low** to **Max**. Each model shows its recommended setting, and you can raise it only when a task calls for it. The Primary model list with Claude Fable 5 selected and the Effort slider set to Medium, described as balanced reasoning for everyday tasks Use the level descriptions as your guide: | Effort | Use it for | | - | - | | **Low** | Fastest responses with minimal reasoning. Routine, well-defined changes where speed matters most. | | **Medium** | Balanced reasoning for everyday tasks. A solid default when the recommended setting doesn't say otherwise. | | **High** | Deeper reasoning for more complex work, such as multi-step features and less familiar territory. | | **Extra High** | Extended reasoning for the hardest problems, such as large refactors and tricky bugs. | | **Max** | Maximum reasoning depth. The highest cost and slowest, when the best possible result is all that matters. | At higher Effort, Agent performs deeper, more deliberate reasoning and invokes its most capable frontier models, improving outcomes on the most complex tasks. Agent applies that extra power selectively. It routes to the more capable models only when a request is genuinely hard, not on every run. Because Agent only reaches for the most capable model on the hardest work, higher Effort adds cost up to \~2x on the toughest tasks. It often adds little or nothing on simpler tasks. Raise it for complex problems, and keep it low for routine changes. ## Compare models **Compare models** runs your next prompt across several models at once, so you can judge results side by side instead of guessing which model fits. It's in beta. To turn it on, expand **Advanced settings** at the bottom of the Agent modes popover and switch on **Compare models**. Comparing models is only supported with Chat. Advanced settings expanded with the Compare models Beta toggle switched off, described as creating up to 4 variants with your next prompt With the toggle on, add up to **4 models** to the comparison. The Agent modes popover with Compare models on and four removable model slots: Kimi K3, Claude Fable 5, Claude Opus 5, and GPT-5.6 Sol. The chat send button reads Compare 4 Your next prompt then runs once per selected model, and each produces its own variant to compare. Each model in a comparison runs a separate agent and uses credits, so a Compare 4 run costs about as much as four single runs. ## Model selection in Build and Design Build and Design keep separate model selections. This page covers selecting a model for Agent builds. To explore visual directions with Design's model selector, see [Explore with different models](/design/explore-with-different-models). ## Related * [Agent modes](/features/agent/agent-modes): Choose between Free Mode, Power Mode, and Max Mode * [Explore with different models](/design/explore-with-different-models): Use models to explore Design directions On Enterprise, account admins use [Enterprise model controls](/features/collaboration/enterprise-model-controls) to control the model providers and models available in each Workspace.