Xen AI Model Guide for cTrader Coding
This guide explains how to choose the right AI model in Xen when building or modifying cTrader tools. Models are grouped into Standard and Advanced tiers, and selecting the right one will directly affect code quality, speed, and cost.
Standard models are designed for speed and efficiency. They work well for simple tasks, small edits, and quick testing. Advanced models provide stronger reasoning and are more reliable when building, converting, debugging, or extending trading systems.
For most coding tasks, DeepSeek Pro is the recommended starting point.
Xen AI – Model Overview Demo
Watch this video for a quick explanation of how Xen AI models work.
Tip: Set the video quality to HD (1080p) for the best viewing experience.
Image Upload Support
Some Xen workflows allow you to upload a screenshot, such as converting indicators or building trading panels and dashboards. When an image is provided, Xen can analyse the layout and generate a working baseline based on what is visible.
Image input is currently supported by:
- Claude Sonnet 4.6
- GPT 5.5 Max
Other models do not support image uploads. If you attempt to use an unsupported model, Xen will prompt you to switch.
Standard Models
Standard models are lower cost and are best used for simple tasks. They are suitable for small edits, basic indicators, and quick experimentation, but they are not designed for full trading systems.
| Model | Cost Tier | Usage |
|---|---|---|
| DeepSeek Flash | Lowest Cost | Fast, low-cost model for simple coding tasks and basic logic. |
| Gemini Flash | Lowest Cost | Suitable for quick fixes and general questions. |
| GPT 5.4 Mini | Low Cost | Useful for small edits and lightweight coding tasks. |
Warning
Standard models may produce incomplete or incorrect code when used for complex trading systems. Use an Advanced model for more reliable results.
DeepSeek Flash
DeepSeek Flash is designed for speed and low cost. It works well for simple coding tasks and quick changes, but it is not suitable for building full trading systems or handling complex logic.
Gemini Flash
Gemini Flash is optimised for rapid responses and minimal cost. It is useful for basic questions and quick fixes, but it is not intended for full strategy builds or advanced development work.
GPT 5.4 Mini
GPT Mini is a lightweight model that handles small code edits and simple updates. It performs well on minor tasks but may struggle with larger or more complex systems.
Advanced Models
Advanced models provide stronger reasoning and better code quality. They are recommended for most development workflows, especially when accuracy and structure are important.
| Model | Cost Tier | Usage |
|---|---|---|
| DeepSeek Pro | Low Cost Advanced | Recommended default for most coding tasks. |
| GPT 5.5 Max | Highest Cost | Best for complex builds and image workflows. |
| GPT 5.3 Codex | Premium | Strong for debugging and code repair. |
| Claude Sonnet 4.6 | Premium | Strong reasoning model with image support. |
DeepSeek Pro
DeepSeek Pro offers strong coding performance at a relatively low cost and is the recommended default for most users. It works well for building cBots and indicators, refactoring code, and handling more complex logic without excessive credit usage.
GPT 5.5 Max
GPT Max is the highest quality model available in Xen and should be used when accuracy is critical.
Warning
GPT Max uses significantly more credit than other models. It should be reserved for complex builds, difficult debugging, large projects, or image-based workflows where maximum quality is required.
GPT 5.3 Codex
GPT Codex is focused on coding and performs well when fixing errors, debugging failed builds, and refining existing code. It is a strong choice when you need structured and reliable changes rather than generating a full system from scratch.
Claude Sonnet 4.6
Claude Sonnet is a reasoning-focused model with support for image input. It is particularly effective for screenshot-based workflows, such as converting indicators or building trading panels from visual layouts, as well as handling complex multi-step logic.
Summary
Choosing the right model depends on the complexity of your task.
- Use DeepSeek Pro as your default for most coding work
- Use Standard models for simple edits and quick testing
- Use GPT Codex for debugging and code repair
- Use Claude Sonnet or GPT Max for image-based workflows
- Use GPT Max only when maximum quality is required due to higher cost
A practical approach is to start with a capable model, test the result, and switch if needed. Over time, you will identify which models work best for your specific workflow.