Reading model release notes on Sunday (as you do), I underlined the same sentence twice, not because I missed it the first time, but because Moonshot admitted the quiet part out loud. -Kimi K3 on Wednesday, 2.8 trillion parameters. -Qwen 3.8 on Sunday, 2.4 trillion. Both measured against Fable 5, the best closed model today. Mid-launch Moonshot conceded that K3 still trails Fable 5 overall. Signals that the market is maturing? 🧵👇(Teana Baker-Taylor)