Smart Router Explained: How Intelligent Model Routing Saves You Credits

# Smart Router Explained: How Intelligent Model Routing Saves You Credits

Every task you send to an AI has a natural complexity level. Routing a simple summarisation request to a premium reasoning model is like hiring a surgeon to change a lightbulb — technically it works, but the cost is absurd. Vincony's Smart Router solves this automatically, analysing each prompt and selecting the cheapest model capable of handling it well. Here is exactly how it works and why it matters for your credit balance.

What Is Smart Router?

Smart Router is a built-in feature on Vincony that intercepts every request before it reaches a model, evaluates the task's requirements, and dispatches it to the most cost-efficient model likely to produce a good result. You never touch a model selector; the routing decision is made in the background before you see a response.

This matters because Vincony's catalogue currently spans 750+ distinct models across 80+ providers. The range is enormous — from compact, fast, cheap models built for single-sentence answers to multi-modal reasoning giants that handle code review, long-document analysis, and nuanced creative work. Without some form of intelligent dispatch, users either over-pay by defaulting to flagship models or under-get by picking cheap models that cannot handle the task.

Smart Router sits between you and all 750+ models and makes that call for you.

How the Routing Logic Works

Smart Router analyses several signals from each incoming prompt:

Task type detection. Is the prompt a simple factual question, a code-generation task, a long-context analysis, or a creative brief? Each category has a different capability floor.

Context length. Short prompts with no attached documents can be handled by smaller models. Long documents, multi-turn conversations, and large code files require models with bigger context windows.

Output format requirements. Requests that need structured JSON, function calls, or multi-step reasoning are routed to models whose fine-tuning supports those output types reliably.

Credit tier matching. The router knows your current plan and balances quality against your available credit budget. A Power plan user running batch jobs benefits from the same logic as a Starter user having a quick conversation — the router optimises for each situation.

The result is that routine requests — drafting a short email, answering a factual question, reformatting a list — consume 1–2 credits instead of the 3–4 credits charged for premium or reasoning models. Over hundreds of daily requests, the savings compound significantly.

Credits by Request Type: A Reference Table

Understanding how Smart Router maps task complexity to credit cost helps you write prompts that work with the system rather than against it.

For media generation (image, video, 3D), Smart Router has less latitude — quality requirements usually dictate a specific model — but for text tasks, the savings are real and consistent.

A Concrete Example: Writing a Cold Email

Suppose you are on the Starter plan ($16.99/month, 750 credits) and you send 30 cold email drafts per day across a sales campaign.

Prompt: "Write a concise cold email introducing our logistics software to a mid-sized e-commerce warehouse manager. Emphasise fast integration and 24/7 support. Keep it under 150 words."

Without Smart Router, a user who manually selects a premium model like Claude Opus 4.5 or GPT-5.2 would spend 3–4 credits per draft. Thirty drafts = up to 120 credits per day. Your 750-credit monthly budget would be exhausted in about six days.

With Smart Router active, that cold email prompt is correctly identified as a straightforward copywriting task — no long context, no complex reasoning, no structured output required. It routes to a capable but efficient model and costs 1–2 credits per draft. Thirty drafts = 30–60 credits per day. Your 750-credit budget now covers the entire month with headroom to spare.

The quality difference for this task is negligible. Smart Router does not cut quality arbitrarily; it routes to the cheapest model that can genuinely handle the work.

When Smart Router Routes Up (Not Down)

Intelligent routing is not always about saving credits — it is about matching capability to need. Smart Router will escalate to a stronger model when it detects signals that a cheaper model would fail:

• Long context. A prompt containing a 50-page contract for legal summarisation routes to a model with a large enough context window. Sending it to a small model that truncates input would produce a bad answer.
• Structured output. A request for a complex JSON schema with nested fields and validation rules routes to a model reliably fine-tuned for function calling.
• Code with dependencies. A debugging task involving a multi-file codebase and specific library versions routes to a model like GPT-5.2 Codex or DeepSeek V3.2, which are optimised for code.
• Reasoning chains. A prompt that explicitly asks the model to think step-by-step through a logical problem triggers routing toward reasoning-capable models.

In these cases, Smart Router intentionally spends more credits because spending fewer would produce a worse result. The goal is not the cheapest outcome — it is the best outcome per credit.

Smart Router vs. Manual Model Selection

Vincony gives you both options. Smart Router is the default and suits the vast majority of use cases. Manual selection via the model browser is available when you have a specific reason to prefer one model — testing a newly released model, matching a previous conversation's style, or running a benchmark comparison.

Compare Chat is a third option: it lets you send the same prompt to two or three models side-by-side to see how their outputs differ. This is the fastest way to decide whether manual selection is ever worth overriding Smart Router for your specific workflow.

Smart Router Across Plans

Smart Router is available on every Vincony plan, including the free tier.

Users on the Free plan benefit especially: 100 credits per month is a tight budget, and Smart Router ensures that every credit goes as far as possible before you need to upgrade.

Bring Your Own Key (BYOK) and Smart Router

If you connect your own API keys via Vincony's BYOK feature, Smart Router continues to work — it routes across both your BYOK providers and Vincony's pooled access. You can also configure Smart Router to prefer your BYOK connections when routing identical-capability models, which can reduce your Vincony credit spend even further.

Frequently Asked Questions

Can I see which model Smart Router chose for my request? Yes. Each response in Vincony shows the model used, including when Smart Router made the selection. You can inspect the routing decision after the fact without needing to configure anything in advance.

Does Smart Router work with all 750+ models in the catalogue? Smart Router draws from a curated subset of well-tested models rather than every model in the catalogue. It evaluates routing candidates for reliability and output consistency before including them in its pool. You can access the full catalogue manually via the model browser at any time.

Will Smart Router ever route my prompt to an older or less capable model? It routes to the cheapest capable model, not the cheapest model regardless of quality. For a given task type, Smart Router has a capability floor — if no cheap model meets it, it escalates. You will not receive a degraded output from a model that cannot handle the task.

Can I disable Smart Router and always use a specific model? Yes. In any chat session, you can open the model selector and choose a specific model. That choice pins the model for that session. Smart Router only applies when no manual selection has been made.

---

Start exploring Smart Router for free — every Vincony account begins with 100 free credits, and you can test intelligent routing across dozens of text tasks before spending a cent. See all plans and get started at /pricing.