Choosing an AI Model

TradingView Xen provides several coding models through one workspace. The selection is remembered and used for the next request. The most advanced option is not automatically the best for every job; coding ability, reasoning, speed, cost and image support can all affect the result.

Use a capable coding model for complex strategies, indicators and conversions. A lower-cost model can suit straightforward builds, explanations and focused modifications. For Pine Editor errors, Xen recommends GPT Codex 5.3 and may switch to it when the dedicated repair task is opened, while still allowing another available model to be selected.

Choosing for the current task

GPT Sol 5.6, GPT 5.6 Luna and the Claude models support indicator reference images. GPT Codex 5.3 is intended for code repair and does not accept images. Xen displays the current restriction beside the upload control so an image is not silently ignored.

Analyse Strategy is intentionally fixed to GPT 5.6 Luna. The application first normalises completed trades and calculates authoritative statistics, then the model interprets that evidence. Changing the ordinary workspace model does not change the analysis model.

Availability and credit use

Models can change as providers release improved versions or retire older ones. Use the names and capability guidance shown inside Xen as the current source of truth.

Credit use depends on the selected model and the number of input and output tokens. Large source files, long conversations and extensive generated code cost more than concise requests. Lower-cost options are identified in the selector. Switching models can help when a provider is slow or when a difficult repair needs a different coding approach, but improving the request is often more effective than repeatedly regenerating unchanged instructions.