A Google Gemini model now has a “dial” to adjust how much it reasons
1 min read
Summary
A new update for Google DeepMind’s top Gemini AI model has added a dial so developers can control how much the model “thinks” when formulating a response, with the added benefit of saving them money and energy.
The trend towards reasoning models that spend longer considering a response and performing tasks such as analyzing code or reviewing multiple documents has resulted in the cost of certain tasks reaching $200 or more.
However, the new dial feature can be adjusted depending on the complexity of the task at hand and help reduce expenses.
Overthinking is a common issue with such models, which can end up worsening the AI’s environmental footprint and jack up costs for developers.
The new dial is not yet available to consumers but is aimed at developers of apps.
More testing and feedback is needed to ascertain when medium or low settings are suitable.