hybrid-llm-opus-critical-haiku-routine-anthropic-only

ResearchClaw pipeline LLM allocation: Opus 4.6 for 8 critical reasoning stages (hypothesis generation, experiment design, literature synthesis, paper writing, peer review simulation), Haiku 4.5 for 15 routine stages (topic init, resource planning, export, formatting). When evaluating cheaper models for routine stages, stay within the Anthropic ecosystem (Haiku 4.5) rather than GPT-4o-mini or Gemini Flash — ACP already delegates LLM calls through Claude Code, so cross-provider mixing adds complexity without benefit.

Related

• researchclaw-hybrid-llm-haiku45-for-routine-stages
• hybrid-llm-model-pattern-opus-haiku-pipeline-stages
• haiku-4-5-for-routine-researchclaw-stages-not-gpt4o-mini
• hybrid-llm-routing-opus-critical-haiku-routine-for-researchc
• autoresearchclaw-hybrid-llm-routing-opus-haiku
• hybrid-llm-config-opus-critical-stages-haiku-routine