Total hallucination. Carrying on a conversation with itself. Any ideas why this happens? It just happened randomly.
2 Likes
Hi @Derek_R ,
Could you please identify the specific Gemini model you are currently operating on? For the purpose of replication, I would also appreciate it if you could share the prompt that generated this response.
Thank you!
I am using 2.5 Pro. Since I am being forced to have at least 20 characters in my reply which does not make sense.
Same problems!
Gemini 2.5 Pro, invents complete nonsense, does not follow the conversation, I need to correct him 5 times (+ insult him 10), before it gives me the right answer.
Something is totally wrong - did they throttle the model or TPU’s, so they spend more computing on the useless 2.5-Flash-Image nonsense?
As soon as we get Qwen3 or DeepSeek with 500k+ token context I go Chinese.