Anthropic released Claude Sonnet 5 on June 30, making it the default model for all free and paid users, with introductory API pricing of $2/$10 per million input/output tokens (rising to $3/$15 after August 31). Sonnet 5 scores 63.2% on SWE-Bench Pro—up from Sonnet 4.6's 58.1% and closing significantly on Opus 4.8's 69.2%—while beating Opus 4.8 on Terminal-Bench 2.1 at 80.4% vs. 74.6%. The model ships with a native 1 million token context window as standard, and testers consistently report it finishes complex multi-step tasks that earlier Sonnets stalled on and checks its own output without being asked.
The launch arrives as Claude Fable 5 (Anthropic's flagship, priced at $10/$50 per million tokens) returned to global availability on July 1 after an 18-day suspension under a Trump-administration export-control order. That regulatory disruption underscored the fragility of global frontier-model access. Sonnet 5 is positioned as the pragmatic choice for most production workloads previously requiring Opus 4.8: for the same use cases (knowledge work, agentic tasks), teams get comparable performance at one-third the cost. The introductory pricing window ends August 31.
For architects: Sonnet 5 is a pricing floor reset. At $2 input, it's now cheaper than many open-source serve-it-yourself alternatives for equivalent quality on knowledge work. Arora's warning (90% token-cost reductions needed) is already being answered by Anthropic's tiering. Teams locked into Opus 4.8 or evaluating migration to lower-cost Chinese models (DeepSeek, GLM 5.2, Kimi) have a new default path: stay within Anthropic's ecosystem, drop to Sonnet 5, and redeploy the savings to more agentic workloads or latency-sensitive tasks.