LLM Switchboard: Cost-Aware Orchestrator That Picks the Right Model for Every Task

The poster subscribes to Google, OpenAI, and Anthropic because each model fills a unique role—Gemini for huge contexts, ChatGPT for quick queries, and Claude for the best prose—but Claude's high token costs, faster token depletion, and frequent degraded service make it impractical as a single replacement despite a strong preference for its outputs.