AI Chats - Choosing and Retrying Models in BromeAI

1. What is an AI Chat in BromeAI?

AI Chats are the main interface for talking with the AI models available in BromeAI (GPT-5, Claude, Gemini, Grok, etc.).

In a chat, you can:

• type your questions or instructions

• choose which model answers

• change model at any time

• retry a response with the same model or a different one

📷 In the screenshots on this page, you will see the model selector, the detailed model sheet, and the retry menu under a message.

2. The model selector

2.1. Where is the model selector?

In an AI Chat, the model in use is shown in a rounded button just above the input area, for example:

Claude Sonnet 4.5 (see the screenshot below)

This button always displays the model currently selected for this chat.

2.2. Opening the model list

To change the model:

1. Open an AI Chat.

2. Click the model button (e.g. Automatic Routing).

3. The model selection window opens (see the screenshot below).

This window contains:

• at the top, a search field

• in the center, the list of models grouped by provider (BromeAI, OpenAI, Anthropic, Google, xAI, etc.)

• on the left, a detailed panel for the selected model (name, description, capabilities)

Some models may be marked:

• New – model recently added

• Premium – model only available with an eligible plan

ℹ️ The list of models depends on your subscription and may evolve over time.

2.3. Choosing a model

1. In the list, click the desired model (for example GPT-5.1, Claude Sonnet 4.5, Grok 4 Fast, etc.).

2. The model is highlighted in the list.

3. The left panel updates with details of the chosen model.

4. Close the window (or simply start typing): the button next to the text area now shows the new model.

All messages sent afterwards in this chat will be handled by this model, until you select another one.

3. Automatic Routing (BromeAI)

In the BromeAI section of the list, you will find the model:

Automatic Routing – BromeAI (see the 2nd screenshot)

3.1. What Automatic Routing does

Automatic Routing is a special model: instead of being a single fixed model, it analyzes each request and automatically routes it to the most suitable model, in order to maximize accuracy and speed.

In the left-hand panel, you will see in particular:

• a description explaining its automatic routing role

• its capabilities (Vision, Reasoning, Document Processing)

Concretely:

• you chat normally with BromeAI

• for each message, BromeAI internally chooses the model that seems most relevant

✅ If you do not know which model to choose, Automatic Routing is generally a good default option.

4. Understanding the model sheet

When a model is selected in the list, the left panel (see the 2nd screenshot for Automatic Routing) displays:

• Model name (e.g. Automatic Routing, GPT-5.1, Claude Sonnet 4.5…)

• Provider (BromeAI, OpenAI, Anthropic, etc.)

• A possible New or Premium badge

• A short description of the model

• The list of its capabilities:

  • 👁️ Vision – can analyze images you upload

  • 🧠 Reasoning – optimized for advanced reasoning

  • 📄 Document Processing – better suited for document processing

This information helps you understand what the model is designed for and choose the one that best fits your use case.

5. Types of models: reasoning, vision, document, etc.

5.1. Reasoning models

Models with the Reasoning capability are better suited when:

• you have a complex problem to analyze step by step

• you want a detailed explanation or a structured plan

• you work on architecture, complex code, or strategy

They are designed to better handle chains of thought.

5.2. Vision models

Models with the Vision capability can understand images:

• screenshots

• photos

• diagrams, mockups, charts, etc.

You can:

• drag and drop an image into the chat

• add an instruction, for example:

  “Explain this chart to me”

  Find the error in this code screenshot”

A model without Vision will not be able to interpret the content of the image.

5.3. Document processing

Models with Document Processing are indicated in the sheet.
They are suitable for:

• summarizing, reorganizing or querying long texts

• extracting structured information (key points, lists, etc.)

6. Retrying a message (Retry) and changing model afterwards

Under each AI message, you have a small button with a circular arrow icon (see the screenshot below).

Clicking it opens the Retry menu.

6.1. Retry same

The first option in the menu is:

Retry same

This action sends the same request to the same model, in the same chat.
It is useful when:

• the response does not satisfy you (too short, too vague, etc.)

• you are hoping for a more precise or more complete variant

6.2. Changing model from the retry menu

Under Retry same, the menu shows the available providers:

• Anthropic

• BromeAI

• DeepSeek

• Google

• Meta

• Moonshot AI

• OpenAI

• xAI

• etc.

By hovering over one of these providers, you can choose a specific model from that provider, then resend the same request with this new model.

This allows you to:

• quickly compare the responses of several models on exactly the same question

• have one AI rewrite or verify the response of another (for example, having one model verify another model’s answer)

7. Vision, reasoning, web… how to choose in practice?

Here are a few simple guidelines:

• You do not know what to choose
  → Use Automatic Routing. BromeAI handles selecting a suitable model for each message.

• You want a complex analysis, a detailed plan, or strategy
  → Choose a model with the Reasoning capability.

• You are asking a question about an image
  → Select a model with Vision, or use Automatic Routing which can choose one.

• You are comparing responses from multiple models
  → Use the Retry menu under a message to switch from one model to another without rewriting your question.

8. Frequently Asked Questions

Does changing model erase my conversation?

Changing model via the button near the input area does not delete the existing chat history.
You simply continue the conversation with a different model.

Why do two models give different answers?

Each model is trained differently and does not have exactly the same strengths.
It is normal to see different styles, formulations, or levels of detail.
The Retry menu is specifically designed so that you can take advantage of this diversity.