STORY · MODELLER_
Moonshot and Alibaba launch powerful MoE models for agentic coding
Moonshot has released Kimi K2.6, a 1 trillion parameter MoE model with 32B active parameters and 256K context window that shows strong results on coding tasks and long-running agentic executions. Alibaba competitor Qwen3.6-Max-Preview and Hermes Agent also demonstrate significant progress in agentic coding and multi-agent orchestration.
WHY IT MATTERS
The launch of these models marks intense Chinese competition for dominance in open agentic coding models, which is becoming increasingly critical for practical AI deployment. Results on benchmarks like SWE-Bench Pro indicate that open-source alternatives are now seriously competing with proprietary solutions.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.