A support operations manager must pick a Copilot model to triage 2,000 multilingual help-center articles in GitHub, needing fast everyday answers but deeper reasoning for complex troubleshooting guides with clear citations. Budget limits premium use. Which model selection approach recognizes considerations for choosing between models, including the unique strengths of each model?

The tiered approach is right because it matches model strengths to task demands: a fast, efficient model can handle high-volume routine answers across 2,000 multilingual GitHub articles quickly and at lower cost, while a reasoning model is reserved for complex troubleshooting guides that need deeper inference and clear citations. This preserves budget while supporting multilingual coverage. Using a fast model for routine multilingual answers and a reasoning model for deeper analysis on complex guides does not explicitly ensure citations, so the answers may lack the verifiable sources needed for support trust and auditability. Using a fast model for all articles and adding manual citation checks controls model costs, but leaves complex troubleshooting underpowered for sustained, multistep reasoning; checking citations does not supply that reasoning capacity. Pairing routine answers with a reasoning model and citations for complex guides adds weekly full reindexing, an unnecessary operational step that can consume resources and work against the budget constraint; freshness alone does not make that added work worthwhile. Exam caveat: multilingual support, GitHub scale, or citations mentioned in isolation is not enough—the choice must balance all the constraints. Practical check: pilot tiered routing on sampled routine and complex articles, verify citation coverage and multilingual accuracy, and monitor premium-token use against budget before rollout.