Skip to content

Model Trial

Most models in Hydra not only provide descriptive information but also offer an interactive trial feature. For example, you can chat with a large language model to directly experience its capabilities.

experience

Trial Entry Points

Hydra provides two convenient entry points for model trial. Choose the one that best suits your needs:

  • Option 1: On the Model Gallery page, select a model you like and click Try to enter the trial interface.
  • Option 2: From the left menu, click Model Trial, select a model type, and begin. Currently supported types are Text Models and Image Models.

Trial Overview

When you first enter the Trial Center, the system will recommend popular high-quality models for you to try. You can:

  • Click on a recommended model to immediately start a chat with it;
  • Or select other models of interest from the Public Models list;
  • If you’ve deployed your own models, you can also try them from the Model Services list.

Model Types

Text Models

The Trial Center supports pre-configured conversational models, as well as your self-deployed models. Multiple online text models can be selected at once.

Image Models

Currently supports the preconfigured image generation model Hidream-l1-Dev, along with user-deployed services. Only one image model can be used in a trial at a time.

Completing a Full Conversation

Just three steps to complete a trial conversation:

  1. On your first visit, choose an API Key. If you don't have one, the platform can auto-generate it for you.
  2. Select a model from the list and click it to begin. If you’ve already added a model, you can skip this step and start directly.
  3. If you're not satisfied with the response, click Refresh to regenerate. Click Copy to copy the generated content.

At the bottom of the conversation, the number of tokens used is displayed, which helps evaluate model performance.

switch

More Conversation Features

Click Clear Context in the upper-right of the chat box to end the current session and reset the context — new messages won’t be influenced by previous ones.

switch

Model Comparison

Click Add Model in the top-right to compare up to three text-generation models side by side.

comparison

Switching Models

Click the Switch Model button in the top information bar to switch to another model.

switch

Model Parameter Settings

Hydra allows users to adjust various model parameters. Different configurations affect the generated response.
Click Model Config in the top bar to adjust settings.

switch

Hover over the ? next to each parameter to see its detailed explanation.

Parameter Description
System System role: Defines the model’s behavior guidelines and background. For example: “You are an AI assistant.”
Temperature Higher values produce more diverse and random outputs; lower values produce more focused and deterministic responses. Recommended to set either this or top_p.
TopP Controls output diversity. Higher values yield richer responses. Recommended to set only one of top_p or temperature.
Max_tokens Maximum number of tokens the model can generate. Set to 0 for no limit. Suggested values: 500–800 for chat, 800–2000 for short text, 2000–3600 for code, 4000+ for long text.
* Required fields are marked with a red asterisk.
Negative Prompt For image generation: specify content you do not want in the output.
Guidance scale Controls how closely the image adheres to the text description. Higher values mean more accurate images; lower values allow more creativity.

Comments