I am not sure how big AI labs are planning to compete with open-weight models. → Kimi K3 is already as capable as Fable → Qwen 3.8, previewing at 2.4T, is second only to Fable 5 and Opus 5 → GLM 5.2 got massive hype (as capable as Opus 4.8) → GLM 5.5 might beat Opus 5 Opus 5 dropped, but not that impressive. And then other labs like MiniMax and DeepSeek can drop something big any time.