researchclaw-uses-haiku-not-gpt4o-mini-for-routine-stages
AutoResearchClaw hybrid LLM config uses Opus 4.6 for 8 critical reasoning stages (hypothesis generation, paper writing, peer review simulation, etc.) and Haiku 4.5 for 15 routine stages. GPT-4o-mini was explicitly rejected in favor of Haiku 4.5 to stay within the Anthropic ecosystem for ACP protocol compatibility and to keep all inference on the Max subscription at zero marginal cost. AutoResearchClaw natively supports primary + fallback model configuration.