The phrase AI assistant modes describes an assistant that splits your work into distinct jobs — chatting, coding, designing, writing, researching — and sends each one to whichever engine handles it best, instead of forcing every request through the same general-purpose model. Ask Mio is built around exactly this idea: five modes, one account, and no model picker for you to think about.
This matters because a single model that is good at everything is usually a compromise at each individual thing. A model tuned for fast conversational replies is not the same model you want writing production code, and neither is ideal for turning a sentence into a finished illustration. Understanding how mode-based routing works helps you get better results and spend less on tasks that do not need a heavyweight model.
What “Modes” Means in Practice
Instead of one chat box that tries to do everything the same way, Ask Mio organizes work into five modes:
- Chat — everyday questions, explanations, translations, quick ideas
- Code — working code with explanations, bug fixes from a pasted error, refactoring
- Design — logos, banners, social posts, illustrations generated from a sentence
- Write — articles, emails, product descriptions, SEO text
- Research — web search with cited sources, plus PDF and document analysis
Picking a mode tells Mio what kind of output you expect, which narrows down which model family is worth using and what tools (web search, code execution, image generation) should be available for that request.
How Routing Actually Works
You never pick a model
Mio routes every request to the best available model for the task automatically. A short factual question gets a fast, cheap model. A long or nuanced explanation, or a request that needs a bigger context window, gets routed to a model built for that. Coding questions go to coding-oriented models, and image requests go to image-generation models. Paid plans unlock access to stronger models across the board, but the routing logic itself stays invisible to you either way.
Why not just let people choose?
Some assistants expose a model dropdown and let you pick GPT-this or Claude-that yourself. That works fine if you already know the tradeoffs between models, but most people do not, and the “wrong” choice usually means paying for capability you didn’t need or getting a worse answer than a better-suited model would have given. Automatic routing removes that decision entirely, at the cost of some control for the rare user who really does want to force a specific model.
The Five Modes, One at a Time
Chat mode
This is the default for quick back-and-forth: explaining a concept, translating a sentence, brainstorming names, or asking something you’d normally search for. It costs the least — a typical reply is 1 point, rising to 3 for longer or more complex answers. Most people spend the majority of their time here simply because most questions in a normal day are quick ones: what does this term mean, how do I phrase this message, what’s a better word for that sentence.
Code mode
Paste an error message and Code mode returns a fix with an explanation, not just corrected code with no context. It also handles refactoring requests and, on the Coding plan, runs in a code execution sandbox with connectors that can take write actions, which matters if you want the assistant to actually run and verify code rather than just describe it. Because coding questions vary enormously in difficulty, points here run 5–10 depending on how much reasoning and context the fix requires, rather than a flat rate.
Design mode
Design mode turns a plain-language description into a logo, banner, social post graphic or illustration. Image generation costs more than text (20 points per image) because image models are more computationally expensive to run, and it is only available from the Design plan upward. A useful habit is to describe composition, mood and color before naming the exact subject, since vague prompts tend to produce vague images regardless of which model generates them.
Write mode
Write mode is tuned for longer-form output: articles, emails, product descriptions and SEO text. It is available starting on the Free plan, alongside Chat, which makes it one of the two modes you can try without paying anything. It differs from Chat mode less in the underlying model and more in how the assistant structures its response — expecting a finished, publishable draft rather than a short conversational answer.
Research mode
Research mode searches the web and cites its sources, and also analyzes PDFs and other documents you upload. This is the mode to use when an answer needs to be current or verifiable, rather than pattern-matched from training data alone. It is the mode most worth using for anything involving recent events, prices, regulations or statistics, precisely because those are the categories where an uncited answer is the least trustworthy.
Points, Context and Why Cost Varies by Mode
The point cost of a request is not arbitrary; it roughly tracks how much computation the underlying model needs. A one-line factual answer needs a small, fast model running for a fraction of a second. A long research answer that reads multiple sources, or a coding fix that has to reason through several files’ worth of context, needs a model with a larger context window and more reasoning steps, which costs more to run and therefore more points. Understanding this is useful practically: if you only need a quick fact, Chat mode is both faster and cheaper than routing the same question through Research mode “just to be safe.”
Modes vs Plans: What You Actually Get
| Plan | Modes included | Points/month | Notes |
|---|---|---|---|
| Free (€0) | Chat, Write | 600 | No images, no file uploads, 30-day history |
| Chat (€5/mo) | Chat, Write, Research | 2,000 | Adds web search, file uploads, memory |
| Coding (€12/mo) | + Code | 6,000 | Code sandbox, connectors with write actions |
| Design (€12/mo) | + Design | 5,000 | 150 images/month included |
| Business (€29/mo) | All five | 15,000 | 50 seats, API access, auto-refill |
A Practical Example
Say you’re preparing a client proposal. You might use Research mode to pull current facts with sources, Write mode to draft the proposal text, Design mode to generate a cover banner, and Chat mode afterward to answer a quick question about formatting. In a single-model assistant, all four of those requests hit the same general model. In Mio, each one is routed to whatever is actually built for that job, and you never had to tell it which model to use — only which mode fits what you’re doing right now, which is usually obvious from the task itself.
Add a second example: a small e-commerce shop preparing for a product launch might use Research mode to check a competitor’s current pricing with sources attached, Design mode to generate a handful of banner variants for testing, Write mode for the product description and launch email, and Code mode to fix a small bug in the checkout page copy script. None of that requires switching accounts or tools — just picking the mode that matches the task in front of you.
When Mode Switching Isn’t Necessary
Not every workflow needs five separate modes. If you mostly ask quick questions and occasionally draft an email, Chat and Write alone — both included on the Free plan — probably cover you. Mode-based routing pays off most clearly once your work spans multiple types of output regularly: someone doing customer support, marketing and light reporting in the same week benefits far more from automatic routing than someone who only ever chats.
Common Mistakes When Using Modes
Defaulting to Chat for everything
It’s tempting to just stay in the mode you opened first. The cost is usually small for a single request, but over a month, routing research questions through Chat mode instead of Research mode means missing citations you might actually need to check later, and routing coding questions through Chat means missing the sandbox execution that would have caught a bug before you copied the code out.
Expecting Design mode to read your mind
A one-word prompt like “logo” will produce something generic in any image tool. Mode-based routing picks the right model family, but you still need to describe style, color and composition for Design mode to produce something usable on the first try.
Ignoring plan limits when picking a mode
Trying to use Code or Design mode on a plan that doesn’t include them will not work the way you expect. Check which modes your plan actually unlocks before building a workflow around one, since upgrading later is simpler than discovering mid-project that a mode you need isn’t available.
Treating every answer as final
Routing to the right model improves the odds of a good answer, but it doesn’t remove the need to check anything a coding fix touches in production, or anything a research answer will be quoted from. Automatic routing changes which model answers you, not whether you should still read the answer critically.
How This Differs From “One Big Model” Assistants
Some competing assistants are built around a single flagship model with optional add-on tools bolted onto the same chat interface. That approach can work well, but it usually means every request, simple or complex, at least starts by going through the same core model before any tool is invoked. Ask Mio’s mode structure decides the model path upfront, based on what you selected, which tends to keep costs more predictable and keeps a small quick question from costing as much as a genuinely hard one. For a closer side-by-side, see how this plays out against a single-model competitor in our comparison of Ask Mio and ChatGPT.
Frequently Asked Questions
Do I have to manually switch modes for every request?
You select a mode when you start a conversation or task, and it stays active for that thread. You are not re-selecting a mode for every single message within the same conversation.
What happens if I use the wrong mode?
Nothing breaks. Asking a coding question in Chat mode will still usually get you code, just without the sandbox execution and connector features Code mode adds. Picking the closest matching mode simply gets you the best-suited tools and model for that job.
Does mode choice affect how many points a request costs?
Yes. A Chat reply typically costs 1–3 points, a coding answer 5–10, and a generated image 20, reflecting the different computational cost of each underlying model.
Can I access all five modes on one plan?
Only the Business plan (29 €/month or 290 €/year) includes all five modes by default; other plans add specific modes as listed in the pricing page.
Does research mode actually cite real sources?
Yes, Research mode performs live web search and returns cited sources alongside its answer, rather than presenting an answer with no way to check where it came from.
Is model routing the same as “auto mode” in other AI apps?
It’s conceptually similar to auto-select features elsewhere, but Mio’s routing spans five distinct task types with different tool access (search, code execution, image generation), not just a choice between two or three chat models.
Will routing choices change over time?
Very likely. Because Mio is not locked to one model vendor, it can swap in stronger or cheaper models behind the scenes as they become available, without changing how you use the five modes.
Can I switch modes mid-conversation?
Starting a new task in a different mode is normal and expected; most people naturally move between modes throughout a session rather than sticking to one mode for an entire workday.
The Bottom Line
Mode-based routing exists to remove a decision you probably do not want to make anyway: which model should handle this. Ask Mio’s five modes map cleanly onto how people actually split their work — chatting, coding, designing, writing, researching — and route each one to a model suited for it, with plan-based access controlling which modes you can use. If you want to see how the modes feel on real tasks, the free plan gives you Chat and Write with no card required.