🚀 #MoonshotAI releases #KimiK26 — #opensource 1T-param MoE model, 32B active params, 262K context & native multimodal (image + video) input🧠 +15% over K2.5 on benchmarks: better long-horizon coding stability, deeper reasoning & stronger instruction following #AI #LLM🤖 #AgentSwarm now scales to 300 parallel sub-agents coordinating 4,000 steps simultaneously — up from 100 agents in K2.5🧵 👇
🚀 #MoonshotAI releases #KimiK26 — #opensource 1T-param MoE model, 32B active params, 262K context & native multimodal (i...