Anthropic has launched Claude Opus 4 and Claude Sonnet 4, marking a leap in hybrid AI models tailored for coding, advanced reasoning, and multi-step agent workflows. These models debuted on May 22, 2025, and are now accessible via Anthropic’s API, Amazon Bedrock, Google Cloud Vertex AI, and integrated into platforms like GitHub Copilot and Snowflake Cortex AI. Opus 4 leads with a 200K context window, excelling at complex, autonomous coding tasks with benchmark-topping performance on SWE-bench (72.5%) and Terminal-bench (43.2%), while sustaining multi-hour sessions without degradation. Sonnet 4 builds on its 3.7 predecessor with improved coding, reasoning, and instruction following, making it ideal for production use where performance and efficiency are balanced.

Key new capabilities include extended thinking with tool use, parallel execution, and robust memory when models are allowed access to local files, enabling better continuity over long tasks. Both models significantly reduce shortcut-seeking behavior compared to prior versions, and introducing “thinking summaries” further optimizes response transparency. Pricing for Opus 4 is set at $15 per million input tokens and $75 per million output tokens, while Sonnet 4 is available at $3/$15 levels.

Industry reactions have highlighted Opus 4’s ability to perform autonomous, hours-long coding sprints and outpace OpenAI’s GPT-4.1 and Google Gemini 2.5 Pro on code-specific benchmarks. Developers and enterprises adopting these models can harness them for everything from background code generation, robust agentic workflows, to complex research work. The updates underscore Anthropic’s push to provide both instant and deep-reasoning capabilities for diverse users, including free users with Sonnet 4 access and enterprise clients leveraging the most powerful features for mission-critical tasks.
The company continues its focus on safety, reliability, and developer-friendly expansion, making Claude 4 models a cornerstone for next-gen AI-driven productivity and collaboration across industries.