Text Models

Text models in Armox are powerful large language models (LLMs) that can generate, analyze, and transform text. Many also support multimodal inputs like images and videos.

Overview

Text models are the foundation of intelligent workflows. Use them to:

• Generate content — Write articles, scripts, marketing copy, and more
• Analyze and reason — Process complex information and provide insights
• Transform text — Rewrite, summarize, translate, or expand content
• Understand media — Analyze images and videos with vision-capable models

Available Text Models

Connection Colors

In the Armox Canvas, text connections use blue handles and edges:

• Input Handle: Blue circle on the left side of nodes
• Output Handle: Blue circle on the right side of nodes
• Connection Edge: Blue line connecting nodes

Common Settings

Most text models share these configuration options:

System Prompt

Set the model's behavior and persona. This is like giving the AI its job description before it starts working.

Max Tokens

Control the maximum length of the response. Higher values allow longer outputs but cost more.

Temperature

Adjust creativity vs. consistency:

• Low (0.0-0.3): Consistent, focused responses
• Medium (0.4-0.7): Balanced creativity
• High (0.8-2.0): More creative, varied outputs

Top P (Nucleus Sampling)

Fine-tune response diversity. Lower values make outputs more deterministic.

Choosing the Right Model

For Speed and Cost

• Llama 3 8B (2 credits) — Fastest, most affordable
• DeepSeek V3.1 (10 credits) — Great balance of speed and capability
• Gemini 2.5 Flash (10 credits) — Fast with vision support

For Quality

• GPT-5.4 (40 credits) — Default frontier model for complex work
• Claude Opus 4.6 (60 credits) — Premium depth for highest-stakes tasks
• Gemini 3.1 Pro (25 credits) — Strong multimodal reasoning with medium thinking default
• GPT-5 (20 credits) — Best for complex reasoning
• Claude 4.5 Sonnet (30 credits) — Excellent for writing and nuance
• Grok 4 (20 credits) — Strong problem-solving

For Vision Tasks

• Gemini 3.1 Pro — Images, videos, and audio with balanced reasoning
• GPT-5.4 — Premium multi-image analysis with configurable effort
• Claude Opus 4.6 — High-quality deep analysis with image support
• Gemini 2.5 Flash — Images and videos, fast processing
• GPT-5 — Multiple images, detailed analysis
• Claude 4.5 Sonnet — Single image with deep understanding

Best Practices

1. Be specific — Clear prompts yield better results
2. Use system prompts — Set context and expectations
3. Iterate — Refine prompts based on outputs
4. Match model to task — Don't overpay for simple tasks

Next Steps

Explore individual model documentation for detailed settings and use cases:

• GPT-5.4
• Gemini 3.1 Pro
• Claude Opus 4.6
• GPT-5
• Gemini 2.5 Flash
• Claude 4.5 Sonnet
• DeepSeek V3.1
• Grok 4
• Llama 3 70B
• Llama 3 8B