Gemini has gotten wildly worse in recent time, almost incapable of answering a question without a search, lengthy "thinking time" for trivial questions that no doubt is intended to mask waiting for inference resources
Yesterday I read that they sold a lot of TPUs to Anthropic as well as renting them out. This is odd, because I've constantly had issues with Google LLMs not having enough resources. Except for AI suggestions in Search, these work reliably.
Yesterday I read that they sold a lot of TPUs to Anthropic as well as renting them out. This is odd, because I've constantly had issues with Google LLMs not having enough resources. Except for AI suggestions in Search, these work reliably.