Anthropic released Claude Sonnet 5 on June 30, 2026, positioning it as the most capable Sonnet-series model to date — one that narrows the performance gap with the company’s Opus flagship while running at significantly lower cost.
Built for autonomous work
Sonnet 5 is designed around agentic, multi-step tasks. The model can plan ahead, browse the web, execute code in a terminal, and work through complex assignments with limited human direction — capabilities that until recently required Opus-class models. Anthropic describes it as the most autonomous Sonnet yet.
On the OSWorld-Verified benchmark, which tests computer-use abilities, Sonnet 5 scored 83.4 percent, up from 78.5 percent for the previous Claude Sonnet 4.6.
Pricing and availability
Introductory API pricing through August 31, 2026 is set at $2 per million input tokens and $10 per million output tokens. After that date, pricing rises to $3 input and $15 output per million tokens — still well below Opus 4.8 rates.
The model is the new default for Anthropic’s Free and Pro subscription tiers and is also available to Max, Team, and Enterprise customers. Developers can access it through the Claude API, Claude Code, Amazon Bedrock, and Microsoft Azure AI Foundry, with Google Cloud Vertex AI support listed as coming soon.
Cost-performance range expands
Early enterprise partners reported success applying Sonnet 5 to software engineering automation, end-to-end workflow execution, and legal research — scenarios where Opus-class results had previously demanded Opus-class cost. Anthropic frames the release as giving developers a wider cost-performance range without forcing a trade-off between capability and budget.