Thinking Machine Lab introduced Inkling, an open-weights model with 975B parameters. Well, that's a huge number. If we talk about where it stands, then no, it's not near the top frontier models we have in the market like Fable 5 or the GPT-5.6 series. But it is the leader among the open-source models we have right now.

Also, today or tomorrow, Moonshot AI is going to launch Kimi K3. And it's expected to be a GPT-5.6 and near-Fable 5 level open-weight model from China. It is also expected to have 2–3T parameters. So yeah, it will be the biggest open-source model from China.

But what do you think about these huge parameter numbers? We treat the numbers like gravity. The bigger the model, the smarter it is.

But what if I ask you why bigger models learn more compared to smaller models? What would you say? More parameters.. more.. brain??? Well, that's not an answer.