Google killed Gemini 3.5 Pro and told you why on a podcast. Pichai told Alex H… · @harden_369741 · TwKit

@harden_369741

Google killed Gemini 3.5 Pro and told you why on a podcast. Pichai told Alex Heath that Argon burned ~25 trillion tokens in its first 72 hours, roughly 3x what Gemini 3 did in the same window. Paid API, Pro and Ultra only. When your flagship eats that much compute in three days, you are rationing, whether you say the word or not. 3.5 Pro was supposed to ship in June. By September Google said it wasn't coming at all, and folded the work into Argon's post-training. Not delayed. Deleted. The part I keep coming back to: Google's own 3.5 Flash was already beating 3.1 Pro on the agentic coding benchmarks Google chose to highlight. Good luck shipping a Pro that loses to your own Flash. So the ladder collapsed into one model and a ration card. 25 trillion tokens in 72 hours says the bet is working. It also says there was no compute left for a second mouth to feed. My read: the tidy Pro/Flash/Ultra lineup is dead. When the mid-tier beats the flagship at the job that matters, you don't get three tiers, you get one model and a waitlist. Every other lab is about to do the same math, they're just quieter about it. Source: https://www.techarcade.io/google-s-newest-ai-burned-through-25-trillion-tokens-in-72-hours-now-it-s-rationing/ #Gemini #GoogleDeepMind #AIModels

曝光 10 · 评论 0 · 点赞 0 · 书签 0