Using Model Switching in Harper Assist

Harper Assist now includes model switching, allowing you to choose the most suitable AI model for the task you are working on, all from within the same conversation.

This means you can switch between models depending on whether you need:

• a fast answer
• deeper reasoning
• or help working with large documents

You can also watch the intro video here:

What is model switching?

Model switching allows you to change the AI model used in your current Harper Assist chat without starting again.

Different models are better suited to different types of work. For example:

• some are optimised for speed
• some are designed for complex reasoning
• and some perform particularly well when working with long files or large documents

This gives you more flexibility and helps you get better results depending on the task.

Available models in Harper Assist

GPT 5.3 for fast, everyday answers

GPT 5.3 is the default model in Harper Assist.

It is best used when you want:

• quick answers
• straightforward drafting support
• general guidance
• fast responses to everyday questions

If speed matters most, GPT 5.3 is usually the right choice.

GPT 5.4 for deeper reasoning and more complex questions

GPT 5.4 is the new reasoning model.

It is designed for tasks that need:

• deeper analysis
• more structured thinking
• comparison of different options
• more careful consideration of implications and risks
• better handling of difficult research questions

This model is especially useful when you are dealing with more complex legal or commercial scenarios and want a more analytical response.

When using GPT 5.4, you can choose between:

• Standard Thinking Effort
• Extended Thinking Effort

Standard Thinking Effort

Use this when you want stronger reasoning without waiting too long for a response.

Extended Thinking Effort

Use this when you want the model to spend more time working through the question in greater depth.

Because GPT 5.4 is doing more reasoning, responses can take longer than GPT 5.3 — especially when Extended Thinking Effort is selected. While it is working, Harper Assist will provide feedback on screen.

GPT 4.1 for long documents and file-heavy tasks

GPT 4.1 is particularly useful when working with large files or long documents.

Use GPT 4.1 when you need help with tasks such as:

• summarising a long witness statement
• reviewing a contract bundle
• extracting key points from uploaded files
• analysing lengthy attachments
• working through larger volumes of text

If your main task involves an uploaded document rather than general Q&A, GPT 4.1 may be the best option.

Why use GPT 5.4 with the Harper James Knowledge Base?

The new reasoning model, GPT 5.4, works particularly well with the Harper James Knowledge Base.

When you ask a question, GPT 5.4 can reason through the request more carefully and perform multiple searches across the Knowledge Base to find the most relevant information. It can then use those results to build a more accurate and better-structured response.

This makes GPT 5.4 especially valuable for:

• more nuanced legal or process questions
• questions where the answer may be spread across several internal knowledge articles
• scenarios where context and comparison matter
• finding more complete answers from Harper James internal guidance

In short, if you want a more thorough answer based on the Harper James Knowledge Base, GPT 5.4 is often the best choice.

When should I use each model?

Use GPT 5.3 when:

• you want a quick answer
• the task is straightforward
• speed is more important than deep analysis
• you are doing everyday drafting or guidance tasks

Use GPT 5.4 when:

• the question is more complex
• you want more structured analysis
• you need to compare options or assess risks
• you want stronger reasoning over Harper James Knowledge Base content
• you are happy to wait longer for a more considered answer

Use GPT 4.1 when:

• you are uploading large files
• you need help with long documents
• you want to summarise, extract, or review substantial amounts of text

Do I need to start a new chat to change model?

No.

One of the main benefits of model switching is that you can change model within the same conversation. There is no need to restart your chat every time your task changes.

For example, you might:

• start with GPT 5.3 for a quick answer
• switch to GPT 5.4 to explore a more complex follow-up question
• then switch to GPT 4.1 to review a long uploaded document

This makes it easier to adapt Harper Assist to the work you are doing.

Things to keep in mind

• GPT 5.3 is usually the fastest option
• GPT 5.4 may take longer, especially with Extended Thinking Effort
• GPT 4.1 is often the best choice for long documents and larger uploads
• the best model depends on the task, so it is worth trying different options

Quick summary

Harper Assist now gives you the flexibility to choose the best model for the job:

• GPT 5.3 fast, everyday answers
• GPT 5.4 deeper reasoning and stronger analysis
• GPT 4.1 large documents and file-heavy tasks

If you are working with the Harper James Knowledge Base, GPT 5.4 is especially effective because it can reason through the question and perform multiple Knowledge Base searches to produce a more accurate answer.

Learn more

Watch the intro video here:
Using Model Switching in Harper Assist

If you have any questions or feedback, please use the Feedback menu in Core.