Massive AI downgrade lately? feels like Gemini went back years in time tbh

Google's Gemini Pro, the premium subscription tier of its flagship AI model, is at the center of a growing user revolt over a perceived and severe performance regression. Subscribers on platforms like Reddit are reporting that the model, which they pay for expecting advanced reasoning, now behaves erratically. Users describe needing to repeat instructions constantly, encountering 'absurd mistakes' on pure logic tasks, and watching the AI 'lose the plot entirely' during multi-prompt workflows. The consensus among frustrated posters is that the model's capability has rolled back years, failing on tasks it previously handled with ease from vague prompts.

The core complaints highlight a breakdown in Gemini Pro's contextual understanding and reasoning—key selling points for a paid tier. Users trying to chain a few prompts together for a complex task find the model forgetting the objective and context, making coherent output 'straight up impossible.' This perceived downgrade is especially galling for professionals and power users who subscribed specifically for reliable, advanced logic. While it's unclear if this is a widespread technical issue, a temporary scaling problem, or an intentional model adjustment, the user experience has significantly degraded. The situation raises questions about the stability and testing of updates to major, paid AI services, where users expect consistent or improving performance, not regression.