Agentic coding benchmarks look great.

I think we will find the market is going to move into omni multi-modal models: text, moe, agentic tasks, voice, images, and video all in one LLM.