Feedback on Gemini 2.0 Flash Thinking Experimental 01-21 and Pro Experimental 02-05 Models

I’m currently working on a TypeScript coding project with a token count of approximately 400K, primarily using the gemini-2.0-flash-thinking-exp-01-21 model. This model has excelled at maintaining context despite the expanding token window. However, I’ve encountered some issues with the “thinking” control; it occasionally fails to display or is completely absent. Overall, though, the model has performed quite well.

Hi @shivhoam,

Welcome to forum, thank you for the feedback.