Benchmarked Kimi K2 LLM. It has done well. DeepSeek V3 beats it but Kimi K2 might be more skilled. Very close performance to Qwen 3 in terms of skills and human alignment. But huge parameter count (1T!).