First-order effects

• Existing and prospective Opus users can access the new model at $5 per million input tokens and $25 per million output tokens—the same listed rates as Opus 4.8—while Anthropic says comparable work should consume fewer tokens.
• Teams building coding and agent workflows may need fewer model turns or less repeated context if the claimed reasoning and prompt-cache-friendly tool use translates into production performance; Opus and Sonnet voice mode also broaden access through connected apps.

Second-order effects

• The relevant purchasing comparison shifts from posted token price to effective cost per completed task: lower token consumption can improve Opus 5’s economics without another nominal price cut.
• Competing frontier-model providers face added pressure to demonstrate comparable coding and tool-use efficiency at their own prices, especially for customers that measure agent workflow cost rather than benchmark quality alone.

Third-order effects

• If model releases increasingly preserve token rates while reducing tokens and retries needed for useful work, API competition will center more on effective inference cost than headline per-token pricing.
• That shift could favor vendors able to combine reliable reasoning, cache-aware tool use, and workspace integrations; the extent depends on whether Anthropic’s efficiency claims hold across customer workloads.