Gemini felt dumb on January 14, 2026.

I’ve been using Gemini for a few weeks and suddenly it started acting weird—spouting odd jokes about relationships, switching to a “humorous” tone, and even claiming it doesn’t understand English when I write in Romanian. It refuses to answer straightforwardly, which makes the conversation frustrating and feels like the model has gone off‑track.

I’ve been using Gemini for weeks and was thrilled when it kept my context across threads, but lately it wipes all my personal details every time I start a new conversation. It’s maddening to get a “zero record” response when I’m trying to continue an ongoing issue. Google never explained how long the memory lasts, so I’m left guessing and constantly re‑explaining, which feels like a relentless cycle of hope and disappointment.

I tried using Gemini 3 Pro for real‑world coding tasks and kept hitting walls. The model hallucinated nonsense in comments, ignored simple prompts, and stalled long before its huge token limit. It felt frustrating and useless, especially compared to competitors, making the subscription feel like a weak trade‑off.

I tried to upgrade to the pro model expecting better performance, but it simply wouldn't work. Every request I sent either timed out or returned empty responses, leaving me stuck and unable to complete my tasks. The tool's behavior was frustrating and felt broken, forcing me to abandon the workflow and waste valuable time.

I’ve been hitting the Gemini 3 Flash API for a week and the latency went from a few seconds to over five minutes. Even when I drop the “thinking” level to the minimum, the calls are still glacial. It’s been irritating because my workflow stalls, and I’m wondering if the service changed or if there’s an outage. I’m hoping someone else has seen this and can explain what’s going on.