Wow, Kimi 2.6 is here! Looking at the benchmark data, it's basically in the same league as Claude Opus 4.6 and GPT-5.4. Probably even stronger than GLM 5.1, and it's a multimodal model too!!