One of the new API features with GPT 5 is the Verbosity parameter, which roughly affects how many tokens are generated, i.e. how wordy the model is, which can either negatively or positively affect output quality depending on the task. It has three modes (low, medium, high).
https://platform.openai.com/docs/guides/latest-model#verbosity
Along with verbosity, there’s a new Minimal setting for reasoning. This allows the model to think just a little bit before answering, which often improves answer quality while keeping response time and token usage low.
https://platform.openai.com/docs/guides/latest-model#minimal-reasoning-effort
Seeing as there’s already a slash command for /reasoning_effort (high/medium/low), there should be Minimal too. When used with models that don’t support that setting, it should use Low or the next lowest reasoning option instead.
There could be a slash command such as /verbosity (low/medium/high), though that would need to be ignored by all non-GPT 5 models.
Ideally these model parameters and slash commands would be unified into a UI for fuller customization of model responses, like something in the Google AI studio.
EDIT: Include Mini and Nano, as they don’t seem to support it at the moment.
