Updated default model definition and all high-token agents (Trade, Research, Auditor) to the new Qwen3.6 abliteration.
- Default + Ops → Grok 4.1 Fast (light & fast) - Trade, Research, Auditor → Qwen3.5-27B on A100 (high token / heavy reasoning) - Updated defaults and added thinking levels where appropriate - Prepares for routing high-load work to local vLLM