Qwen3.6 35B (3B active)

User runs Qwen3.6-35B-A3B-4bit on an M3 Max 128GB for production sub-agent delegations. User also mentions GLM-5.1 for orchestration. User is considering building a 5090 rig.