Google has introduced a new 'thinking level' feature for its Gemini 4 Neon model, allowing users to control the depth of the model's reasoning. The feature offers four settings: low, medium, high (default), and xhigh. Google recommends using 'high' for most complex tasks and only switching to 'xhigh' for challenging, multi-step problems, as 'xhigh' consumes more tokens and increases latency. The 'thinking level' can be configured differently depending on whether the Interactions API or the generateContent API is used. AI
IMPACT Allows developers to fine-tune reasoning depth and cost for specific tasks, balancing performance with resource consumption.
RANK_REASON The article details a specific feature and its configuration within an existing AI model, rather than a new model release or significant research breakthrough.
- Gemini 3 Flash
- Gemini 4
- Gemini 4 Argon
- Gemini 4 Hydrogen
- Gemini 4 Neon
- Gemini API
- generateContent
- Interactions API
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →