Run Prompts with AI Models
Run up to six connected models and compare outputs, timing, tokens, and cost estimates.
Run prompts with AI models
PromptBranch can run one prompt with up to six connected models at once, so you can compare answers instead of guessing which model or prompt version works best.
Use variables
Write reusable input fields as {{variable_name}} in a prompt. Selecting Run asks for a value for every variable it finds. PromptBranch remembers those values for the prompt so you can run it again without re-entering them.
Run and compare
- Choose one or more models in the model picker next to Run.
- Select Run and fill in any variables.
- Watch each model's result as it arrives.
Open the Results tab and select a run group to compare its outputs. Each result shows its status, response, elapsed time, token usage when the provider reports it, and an estimated cost when pricing is available. You can cancel running requests; a provider may still charge for work completed before the cancellation.
Cost is an estimate, not a bill. Local models and models without published pricing show no estimate.
Next: evaluate and judge results.