Past April of 2026, all open-source LLMs have been in the 0.5T+ terrirory: MiniMax M3, Kimi-2.7-Code - now Kimi-3 (2.8T), GLM-5.2, Inkling by ThinkingMachines is also a 1T model and perhaps some more models I forget now. If chinese labs find it more profitable (I would not accuse them since it takes
