Anthropic released Claude Opus 5, a model cheaper and less restrictive than Fable 5, with effort-toggle controls (low/medium/high) letting users trade cost against output quality per query. Teams running coding and agentic workloads at scale gain direct pricing levers without model switching.

Opus 5 introduces a cost-capability toggle addressing enterprise customer concern over AI infrastructure spend, shipping with configurable effort levels. This shift signals vendor recognition that billing-per-query economics now drive adoption friction in production workflows.

OpenAI's GPT-5.6 tiered lineup spans $1–$30 per million tokens (Luna to Sol), with Sol at feature parity to Opus 5 at $5/$30. Multi-tier stacking now standard across OpenAI and Anthropic, forcing builders to benchmark latency, quality, and cost across eight-plus production models.

Anthropic extended introductory Claude Opus 5 pricing ($2/$10 per million tokens) through August 31, then steps to $3/$15 standard rates. Discount window and pricing structure mirror competitive tactics seen with Sonnet 5, signaling aggressive acquisition pricing on enterprise contracts before margin normalization.