LLM
The LLM node generates text from a prompt, using a language model you choose. Use it to write scripts, captions, ideas, or any other text your workflow needs.
Inputs and outputs
Inputs (all optional): Image, Texts, Prompt, and a generic Flow input for connecting from earlier nodes.
You can connect several nodes to Image. The model gets every picture they hold, including every image in a Media Library and every image an Image Generation node made, in the order they're wired.
Output: a single Text handle.
Writing a prompt
Type your prompt directly into the box at the bottom of the card ("Describe what to generate..."). If the Prompt input is already connected from another node, this box is disabled and shows "Prompt connected from input" instead.
The generated text appears in a scrollable area at the top of the card. Click Copy to copy it to your clipboard.
Choosing a model
Click the model picker in the settings bar to search and choose a model. Type to filter, and results are grouped by provider, with prices shown where available.
The default model is GPT-5.6 Luna.
The node runs on Atlas Cloud, and you choose from a fixed list of GPT, DeepSeek, Grok, Qwen, Kimi, GLM, and Gemini models. Some models read text only: if you connect an image to one of them, the run stops before anything is charged and tells you to pick a model that takes images. A Gemini model can take up to 20 MB of images in one run: with more than that, the run also stops before anything is charged and tells you both sizes.
Running the node
Select the node to reveal its Run button in the top right corner of the card. What happens when you click it depends on what's connected:
- A standalone node with no inputs runs itself. Click Run again to also run everything connected after it.
- If its inputs already have results, running it also runs everything downstream.
- If its inputs aren't ready yet, you'll be asked to choose between running the whole workflow, or running everything up to and including this node.
Click the Inputs / Outputs button (an eye icon) at any time to see exactly what data went into this node and what came out.