Anthropic launches Claude Sonnet 5 as a cheaper way to run agents

The more pointed detail buried in this launch is that Anthropic is effectively cannibalizing Opus adoption from within its own lineup, betting that a cheaper, capable-enough model will capture more total agent volume than a premium model captures in margin. That is a deliberate trade-off, not a side effect.

Modelwire has no prior coverage to anchor this to directly, so it sits in a broader context worth naming: the mid-tier inference price war that has been building across the major labs throughout 2025 and into 2026. The pattern is consistent, OpenAI, Google, and now Anthropic have each used a second-tier model release to pressure competitors on cost while preserving a flagship for prestige benchmarks. Sonnet 5 fits that playbook precisely. What is less clear from available reporting is whether Anthropic's safety guardrail improvements are independently verified or self-reported, which matters considerably for enterprise buyers making procurement decisions.

Watch whether OpenAI responds with a GPT-5.5 price cut or a new mid-tier SKU within the next 60 days. A reactive pricing move would confirm that Sonnet 5 landed with enough enterprise traction to threaten GPT-5.5 positioning, not just generate press coverage.