coding benchmarks

Xiaomi has released the MiMo-v2.5 Family model weights, which demonstrate strong performance in coding and agent benchmarks. The update highlights advancements in AI capabilities, particularly in technical and autonomous task execution.

Alibaba released Qwen3.6-Max-Preview, its most powerful AI model to date, which outperformed six major coding benchmarks and showed improvements in world knowledge and instruction following compared to its predecessor.