> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gosutox.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Reasoning Modes: Flash vs. Think

> Master the dual cognitive engines: Sub-second instantaneous Flash Mode vs. Deep multi-step Think Mode with transparent reasoning traces.

## The Dual-Engine Cognitive Paradigm

Enterprise intelligence requires two fundamentally different operational velocities: **instant conversational agility** for rapid day-to-day execution, and **deep multi-step deliberation** for high-stakes strategic decisions.

AXON Agent provides two dedicated reasoning engines, toggled with a single click in the input toolbar:

```mermaid theme={null}
flowchart TD
    Prompt["💬 User Strategic Prompt"] --> Switch{"⚙️ Reasoning Mode Switch"}
    
    Switch -->|Flash Mode| Fast["⚡ Sub-Second Direct Stream"]
    Switch -->|Think Mode (Pro)| Deep["🧠 Multi-Step Deep Deliberation"]
    
    Fast --> Output1["📄 Rapid Operational Response"]
    
    Deep --> Trace["💭 Live Collapsible Thinking Block\n(Progress Bar • Elapsed Timer • Step Validation)"]
    Trace --> Output2["📊 Comprehensive Boardroom Deliverable"]
```

***

## Flash vs. Think: Side-by-Side Comparison

| Operational Dimension  | ⚡ Flash Mode (Default)                   | 🛡️ Think Mode (Pro Reasoning)                     |
| :--------------------- | :--------------------------------------- | :------------------------------------------------- |
| **Response Latency**   | Sub-second (\< 800ms initial token)      | 3 – 8 seconds of deep autonomous reflection        |
| **Cognitive Approach** | Direct, highly focused linear synthesis  | Multi-hypothesis validation & counter-analysis     |
| **Thinking Trace**     | Streamlined / Minimal trace              | Full collapsible live progress bar & reasoning log |
| **Token Efficiency**   | Ultra-lean token consumption             | High-depth token allocation for complex logic      |
| **Primary Use Cases**  | Live market lookups, drafting, summaries | M\&A due diligence, legal audits, complex modeling |

***

## 1. Flash Mode: High-Velocity Execution

**Flash Mode** is engineered for maximum throughput. It bypasses extended recursive reflection loops to deliver immediate, highly structured answers with zero latency lag.

<CardGroup cols={2}>
  <Card title="Instant API Lookups" icon="bolt">
    Query live stock quotes, FX exchange rates, or single SEC/DART regulatory disclosures in real time.
  </Card>

  <Card title="Document Summaries & Edits" icon="file-lines">
    Quickly distill uploaded PDF notes, reformat text, translate languages, or polish executive drafts.
  </Card>
</CardGroup>

### Concrete "Aha!" Scenario: Rapid Market Check

> **Prompt**: *"What is the latest 10-Q filing date for Apple (AAPL), and what was the reported operating margin?"*
>
> **Flash Execution**: Instantly activates the **SEC EDGAR Connector**, extracts the exact metrics from the latest filing, and streams the verified answer within 600ms—complete with a clickable `[ref:SEC EDGAR: AAPL]` source badge.

***

## 2. Think Mode: Deep Multi-Step Deliberation

**Think Mode (Pro)** unlocks AXON's full cognitive power for ambiguous, multi-variable challenges where hasty answers lead to costly corporate errors.

<Accordion defaultOpen title="🧠 Thinking Process (100%)  |  2.4s" icon="brain">
  <Steps>
    <Step title="Claim Decomposition">
      Decomposing patent claims in Section 4.2 of attached patent PDF...
    </Step>

    <Step title="Regulatory Cross-Referencing">
      Cross-referencing active clearance status against FDA 510(k) approval database...
    </Step>

    <Step title="Conflict Detection">
      Identifying potential indemnification overlap with existing European patents...
    </Step>

    <Step title="Strategic Synthesis">
      Formulating executive risk mitigation matrix for Document Studio deliverable...
    </Step>
  </Steps>
</Accordion>

### The Live Thinking Block Features

* **Real-Time Progress Bar**: Visual progress indicator tracking cognitive convergence from 0% to 100%.
* **Live Elapsed Timer**: Precise stopwatch (`0.8s`, `2.4s`, `4.1s`) showing exact deliberation duration.
* **Auto-Scrolling Reasoning Trace**: Transparently observe how the agent evaluates hypotheses, detects contradictory data points, and verifies logic before committing to the final answer.
* **Smart Auto-Collapse**: The thinking panel automatically collapses once deliverable streaming begins, keeping your workspace clean while remaining fully expandable on-demand.

### Concrete "Aha!" Scenario: Complex Legal & Financial Risk Audit

> **Prompt**: *"Cross-examine our attached Series B Shareholder Agreement (PDF) against the uploaded Cap Table (XLSX). Highlight any liquidation preference overhangs or anti-dilution triggers."*
>
> **Think Execution**: The agent spends 4.2 seconds systematically parsing the liquidation waterfall formulas in the spreadsheet, cross-referencing clause numbers in the legal PDF, testing edge cases, and generating a pixel-perfect boardroom risk audit table in Document Studio.

***

## How to Toggle Reasoning Modes

You can switch between Flash and Think Mode at any time without resetting your conversation history:

1. **Input Toolbar Toggle**: Click the **⚡ FLASH** or **🛡️ PRO** pill button directly inside the central input area.
2. **Right Context Tools Panel**: Select your desired reasoning mode from the **Context Configuration** sidebar.
3. **Turn-by-Turn Flexibility**: Use **Flash Mode** for introductory questions, switch to **Think Mode** for the core analytical challenge, and switch back to **Flash Mode** for final formatting adjustments.

***

## Mathematical Token Billing Parity

AXON maintains strict mathematical billing transparency across both modes:

* In **Flash Mode**, token deductions reflect strictly the raw input prompt and generated answer.
* In **Think Mode**, transparent token accounting covers both the deep reasoning steps and the synthesized output, visible in your real-time session telemetry.
* **Zero Token Waste**: AXON never runs infinite uncontrolled retry loops. All thinking steps are bounded by deterministic safety ceilings.

***

## Next Steps

<CardGroup cols={2}>
  <Card title="1-Turn Tool Symphony" icon="sliders" href="/agent/tools-orchestration">
    Learn how to bind Brains, live APIs, and up to 10 data files simultaneously in either reasoning mode.
  </Card>

  <Card title="Code & HTML Artifact Generation" icon="code" href="/agent/code-generation-html">
    Explore how Think Mode autonomously generates standalone interactive HTML tools and visual calculators.
  </Card>
</CardGroup>