OpenAI launches GPT-6.1 Terra with 50% lower standard API price

The pricing structure is the real story here: a model positioned at one-tenth the per-token cost of a flagship while matching it on an office-software benchmark reframes the competitive axis from raw capability toward cost-per-completed-task. That matters most for teams building agentic workflows at volume, where inference spend scales with every screen interaction rather than every prompt. The speed claim compounds this, since faster completion cuts both token consumption and latency in the same deployment. Whether this pricing holds as usage scales, and how rivals respond on the cost-per-task metric rather than benchmark scores, is the open question worth watching.