Armox
    Armox Academy 📚
    KI-Modelle ReferenzText-ModelleDeepSeek V3.1

    DeepSeek V3.1

    DeepSeek V3.1 ist ein starkes Reasoning-Model mit sehr gutem Preis-Leistungs-Verhältnis — inklusive optionalem Enhanced Thinking Mode für komplexe Probleme.

    Overview

    PropertyValue
    ProviderDeepSeek
    Cost10 credits
    ModalityText
    Vision❌ No
    Prompt RequiredYes

    What It's Best For

    • Cost-effective reasoning — Gute Resultate zu etwa halbem Preis im Vergleich zu Premium Models
    • Technical tasks — Code generation, debugging, analysis
    • Structured outputs — JSON, Tabellen und formatierte Daten
    • General assistance — Vielseitig über viele Use Cases
    • Enhanced reasoning — Optionaler Thinking Mode für komplexe Probleme

    Inputs

    Prompt (Required)

    Der Haupt-Textinput, der beschreibt, was das Model tun soll.

    Connection Color: 🟡 Yellow

    Configuration

    Thinking Mode

    Type: Select

    Enhanced reasoning für komplexe Probleme aktivieren.

    OptionDescription
    NoneStandard fast responses
    Medium (Enhanced Reasoning)Deeper analysis, step-by-step thinking

    Default: None

    When to use Medium:

    • Multi-step math problems
    • Complex logic puzzles
    • Detailed code analysis
    • Strategic planning tasks

    Max Tokens

    Type: Slider
    Range: 1 - 4,096
    Default: 1,024

    Maximale Token-Anzahl.

    Temperature

    Type: Slider
    Range: 0 - 2
    Default: 0.1

    Steuert Randomness. Der niedrige Default (0.1) macht DeepSeek standardmäßig deterministischer.

    Top P

    Type: Slider
    Range: 0 - 1
    Default: 1

    Nucleus sampling für Response Diversity.

    Presence Penalty

    Type: Slider
    Range: -2 to 2
    Default: 0

    Penalisiert Tokens, die schon vorkamen. Positive Werte fördern neue Themen.

    Frequency Penalty

    Type: Slider
    Range: -2 to 2
    Default: 0

    Penalisiert Tokens nach Häufigkeit. Positive Werte reduzieren Wiederholungen.

    Output

    Type: Text
    Connection Color: 🟡 Yellow

    Use Cases

    Code Generation

    Write a Python function that validates email addresses using regex. 
    Include error handling and docstrings.
    

    Data Transformation

    Convert this CSV data into a properly formatted JSON structure 
    with nested objects for related fields.
    

    Technical Documentation

    Write API documentation for this endpoint including request/response 
    examples, error codes, and rate limiting information.
    

    Enhanced Reasoning Task

    With Thinking Mode: Medium

    A company has three products. Product A costs $10 and sells 100 units. 
    Product B costs $15 and sells 80 units. Product C costs $20 and sells 50 units.
    If they want to increase total revenue by 20% while keeping prices fixed, 
    what's the minimum increase in units needed for each product?
    

    Tips for Best Results

    1. Use low temperature — DeepSeek funktioniert gut mit deterministischen Settings
    2. Enable thinking for complex tasks — Lohnt sich für komplexe Aufgaben
    3. Be structured — Klare, organisierte Prompts liefern bessere Outputs
    4. Leverage for technical work — Exzellent für Code und Daten
    5. Iterate quickly — Geringe Kosten erlauben mehr Experimentieren

    Cost Comparison

    ModelCostBest For
    Llama 3 8B2 creditsSimple tasks
    DeepSeek V3.110 creditsBalanced performance
    Gemini 2.5 Flash10 creditsMultimodal tasks
    GPT-520 creditsMaximum quality
    Claude 4.5 Sonnet30 creditsLong-form writing

    DeepSeek V3.1 bietet sehr gutes Value für Text-only Tasks ohne Vision Requirements.