Tools

AI-generated text

llm-gemini update corrects reasoning levels for Gemini 4 models

Version 0.38.1 of the llm-gemini plugin corrects the available reasoning levels for three Gemini 4 models.

llm-gemini update corrects reasoning levels for Gemini 4 models

Version 0.38.1 of the llm-gemini plugin corrects the reasoning levels available for the Gemini 4 models. When the models were added previously, support for the xhigh level on Gemini 4 Hydrogen and Gemini 4 Neon was overlooked; both also support low, medium and high. Gemini 4 Carbon remains limited to those first three levels.

According to the release author, the xhigh setting makes a significant difference for Gemini 4 Hydrogen. In an updated model comparison, Hydrogen used 5,311 output tokens at the high setting, at a cost of 3.99 cents. At xhigh, the same model used 31,477 output tokens and cost 23.61 cents.

The author assessed the pelican produced at the higher setting as competitive with Gemini 4 Neon's result. The example also shows that, for this task, the higher reasoning level substantially increased token usage and cost.