Learn how to use Model Assistant, Models, and Playground to select models, review model details, and test responses before integrating the Model Inference API.
Before you write integration code, you can use the console to get model recommendations, narrow down the catalog, and test model behavior.
Describe your task in Model Assistant on the Home page to get model and configuration suggestions.
Browse and compare models on the Models page, then review model details.
If you are not sure which model to start with, describe your task in Model Assistant on the Home page. Model Assistant recommends suitable models and configuration suggestions based on your goal, and asks follow-up questions when more constraints are needed. You can also start from common scenarios such as document summarization, coding assistance, and text classification.Open the Models page to filter models by organization, category, context window, price, and other dimensions.On each model card, you can review key information such as model name, context window, and price. Open View Details to get additional information, including invocation examples.
The platform provides copy-ready curl and Python examples to help you call models quickly. When calling the API, use the model name exactly as shown on the model details page.
On a model card, choose Try Now to open Playground with the selected model. You can enter prompts directly, observe responses, and compare parameter behavior before integrating the model into your application.
Some model categories, such as embedding models, are not supported in Playground. Call them through the API instead.
In Playground, you can:
Switch models from the model selector.
Pick a preset, such as Balanced, Creative, or Precise, or use a custom configuration.
Use a system prompt to set the assistant’s role, tone, or constraints.
Tune parameters such as Temperature, Top P, Top K, Max tokens, and Frequency penalty.
View the context window and current session usage.
Reset or save the current configuration as needed.
In Playground, Max tokens corresponds to the API parameter max_completion_tokens. For reasoning models, it limits the combined total of reasoning tokens and visible output tokens. For details, see Token Usage and Billing.
Parameters a model does not support are disabled automatically, so only valid settings are sent. Each user’s configuration is saved per model.
On the Home page, you can view usage overview data, including token consumption, API requests, throughput trends, Top Models, Top API Keys, and member usage, to quickly understand the main sources of traffic. Member usage is visible only to administrators.To inspect request details and aggregated metrics, see View Usage and View Metrics.
