AI researcher Simon Willison observed a significant difference in output and cost when using Google's Gemini 4 Hydrogen model with different thinking effort settings. The "xhigh" setting, while more expensive at 23.61 cents and using 31,477 output tokens, produced vastly different results compared to the "high" setting, which cost 3.99 cents and used 5,311 output tokens. AI
IMPACT Highlights the trade-offs between AI model performance, cost, and output quality, influencing user choices and development priorities.
RANK_REASON Commentary on an observed difference in AI model performance and cost.
Read on Bluesky Jetstream — AI desk →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →