Claude Sonnet 5's introductory pricing window closes August 31 as frontier prices keep compressing
Anthropic's Sonnet 5, shipped June 30 at an introductory $2/$10 per million tokens, reverts to $3/$15 at the end of August. Even the list price is a marker of how far frontier pricing has fallen — and the introductory discount shows labs now fight for workloads the way clouds fight for tenants.
An introductory price on a frontier model is a tell. Discount windows are what vendors use when customers have credible alternatives — and with Opus 5 at $5/$25, GPT-5.6-family cuts of 80%, and MIT-licensed open weights within reach of frontier coding scores, alternatives are exactly what buyers have. Sonnet 5's $2/$10 window is Anthropic pricing against churn, not against cost.
The compression is structural, not promotional. Every serious lab has now cut prices or shipped a cheaper tier in 2026, and open-weight releases keep resetting the floor: when a 13-billion-active-parameter open model handles most agentic work, the premium a closed model can charge is capped by the gap it maintains. That gap is measured in months and shrinking.
For builders the practical takeaway is that model cost is becoming a routing decision, not a procurement decision. Teams increasingly route easy calls to cheap or open models and reserve frontier tokens for the hard tail — which is why the labs' real competition is for the default slot in the router, and why introductory windows like this one exist at all.
Spectrum AI Lab — AI coding tools pricing compared, 2026 → · Developers Digest — AI coding tools pricing comparison 2026 →