Voxta Concepts
How the pieces work — models, prompts, reasoning, delivery — rather than what each screen does.
Background for getting more out of Voxta. These pages explain how something works and why it is set up that way. For what a screen does, see the Interface guide; for a specific provider, the Services catalog.
Models
Large language models
Picking a model, what sizes mean, quantization formats.
LLM parameters
Temperature, top-p, repetition penalty and the rest.
Renting GPUs with RunPod
When you need more VRAM than you have.
Prompts
Prompt formatting
The wire format a given model expects, and when auto-detect gets it wrong.
Prompt templates
Shaping the content of the prompt, as opposed to its wrapper.
Thinking models
Reasoning effort, thinking budgets, and telling whether a model actually thought.