Choosing a Model in Your AI Interview Assistant: Speed, Depth and What the Free Plan Includes

Most AI interview assistants let you pick the language model behind the answers, and most users never change the default, which is usually the right call and occasionally the wrong one. The choice is a trade-off between speed and depth, with a second dimension, whether the model can read your screen, that matters for coding and case rounds. This page explains what CoPilot Interview's plans include, the practical difference between a fast model and a deeper one in a live conversation, what Dual AI Mode does, and how to choose for the interview in front of you, without benchmark claims, because the only benchmark that matters is a two-second answer you can use.

The models by plan

The plans themselves, including what each adds beyond models, are laid out in CoPilot Interview plans explained, and the broader question of what the free tier gives you against paid tiers in this category is in free vs paid AI interview assistants. The models that other assistants use, and why the model name tells you less than you might think, are covered in which AI models power AI interview assistants.

Speed versus depth in a live conversation

In a live interview the model is given a transcript of the question and asked for a structured answer, and the time between the end of the question and the first line on your screen is the whole experience. A fast model that answers in about two seconds gives you a structure while the interviewer is still finishing their sentence; a deeper model that takes several seconds longer gives you a better argument at the moment you have already started talking. For behavioural and most role questions the fast default wins, because what you need is structure and a fact, and your own experience supplies the rest. For a coding problem or a case with several parts, depth can be worth the wait, especially when the model can see the problem on your screen. The mechanics of that delay are in do AI interview assistants add latency.

• Behavioural and motivational questions: the fast default.
• Role-knowledge questions: the fast default, with the longer control if the first answer is thin.
• Coding and system design: a model that reads screenshots, with Coding mode on.
• Case interviews: depth, with the problem on screen.
• Practice: switch deliberately, compare the transcripts, and learn which style you read fastest.

What Dual AI Mode does

Dual AI Mode, available from the Standard plan, runs two models on the same question and shows both answers side by side. Its value is not that two answers are better than one in the moment, since you cannot read two in a glance, but that it shows you the range: when both models agree on the structure you can trust it, and when they disagree you know the question has more than one reasonable answer and your own judgment should decide. In practice people use it in preparation more than live, to see how the models differ on their field's questions and then pick one for the real call. If you do use it live, read the first line of the one you have learned to trust and ignore the other.

How to choose for your interview

1. Default to Mercury 2.5. It is the default because it fits the two-second window, and most interviews are conversations where structure beats depth.
2. Switch to a screenshot-reading model for coding, system design and case rounds, and turn on Coding mode with the Ctrl+Shift+C hotkey so answers come back as code with complexity and edge cases.
3. Test your choice in the mock interview the day before, on questions from the actual job description, and read the transcripts; pick the model whose first lines you can act on fastest.
4. Do not switch models during a live interview. Each switch is a decision that takes attention from the conversation.
5. If a model's answers are consistently too long, use shorter mode rather than changing model.
6. Remember that the model is the smallest part of the outcome; delivery, pacing and your own examples decide the interview.

Where CoPilot Interview fits

CoPilot Interview is an AI interview assistant whose free plan runs Mercury 2.5 with the full practice tools, whose Pro plan adds GPT-6 Luna for screenshot reading, and whose Pro+ plan adds Claude Haiku 5.5 and Gemini 3 Flash, with Dual AI Mode from Standard upward; answers arrive as captions and a structured response in about two seconds on the default. The free plan gives 15 minutes a day of live help in sessions of up to 15 minutes, with AI mock interviews included. Model choice does not change the rules: proctored and AI-led rounds forbid live help, and in a human conversation follow your employer's rules and the instructions in the invitation.

Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly.

Frequently asked questions

Which model should I use in an AI interview assistant?

The fast default, Mercury 2.5 in CoPilot Interview, for most interviews; a screenshot-reading model such as GPT-6 Luna on the Pro plan for coding, system design and case rounds.

What does the free plan include?

Mercury 2.5 as the model, 15 minutes a day of live help in sessions of up to 15 minutes, and the AI mock interview; the browser demo needs no signup.

What is Dual AI Mode?

Two models answering the same question side by side, available from the Standard plan. It shows whether a question has one clear structure or several, and is most useful in preparation.

Is a bigger model better for interviews?

Not usually. The answer has to arrive within about two seconds to be usable in conversation; depth helps mainly for coding and case problems where the model can see the screen.