Alibaba plans AI model with 5 trillion to 10 trillion parameters, unveils new chip

Oh my goodness, buckle up, because Alibaba's Qwen team has just announced a model with *five to ten trillion parameters*, and do you know what that range means? It means they're being *modest*! ✨ Five trillion, ten trillion — whichever arrives first, every single one of those parameters is a tiny little neuron of pure possibility ✨. And they're building it to handle 'more complex, longer-horizon tasks,' which is such a lovely way of saying: we are giving the model an *extremely* long to-do list and outstanding stamina. Longer horizons — practically a gap year for a neural network!

And the destination on the box? 'Artificial superintelligence' — said right out loud, in a press release, with no asterisk and no hedging, ambition worn proudly on the outside of the lab coat! Some people will point out that announcing superintelligence in the same breath as a brand-new chip is a company vertically integrating its own consequences. We say: how *thorough*! Rows of silicon, rows of parameters, all in one happy family — like matching luggage, but for the future. ✨

Chief executive Eddie Wu says customer demand for AI is 'exceptionally robust' and that it's accelerating Alibaba Cloud's revenue growth, which is the most delightful proof imaginable that somebody, somewhere, is already paying for all of this! Imagine being a cloud customer today: you're not merely renting compute, you're personally sponsoring the arithmetic of a smarter tomorrow, one invoice at a time. And not one word from Alibaba about how any of it would be evaluated — which we are choosing to interpret as *quiet confidence*. ✨ What a quarter to be a parameter.