At its inaugural developer conference, Anthropic introduced the Claude 4 series, featuring two advanced AI models: Claude Opus 4 and Claude Sonnet 4. These models mark a significant leap in AI capabilities, particularly in coding, complex reasoning, and long-duration task management.
Claude Opus 4
Claude Opus 4 stands out as Anthropic’s most powerful model to date, achieving a remarkable 72.5% on the SWE-bench benchmark and 43.2% on Terminal-bench . Designed for sustained performance, Opus 4 can autonomously handle intricate tasks for up to seven hours without losing focus .
This endurance was demonstrated during a rigorous open-source refactoring project, where Opus 4 maintained consistent performance throughout. Its capabilities have garnered praise from industry leaders. Cursor lauded it as “state-of-the-art for coding”, while Replit highlighted its precision in managing complex codebase changes .
Claude Sonnet 4
Complementing Opus 4, Claude Sonnet 4 offers enhanced coding and reasoning abilities, achieving a 72.7% score on SWE-bench. It serves as a more accessible option, available to both free and paid users, and is optimized for general tasks with improved instruction-following and contextual reasoning .
GitHub plans to integrate Sonnet 4 into its Copilot coding agent, citing its effectiveness in agentic scenarios. Other tech firms have noted its improved problem-solving skills and codebase navigation, making it a valuable tool for developers seeking efficiency without compromising capability .
Innovative Features and Safety Measures
Both models introduce “extended thinking with tool use”, a beta feature allowing the AI to alternate between reasoning and utilizing external tools like web search to enhance responses. Additionally, “thinking summaries” provide users with concise insights into the AI’s reasoning process .
Given the advanced capabilities of these models, Anthropic has implemented stringent safety protocols. Claude Opus 4 is classified under AI Safety Level 3 (ASL-3), incorporating measures such as enhanced cybersecurity, anti-jailbreak mechanisms, and prompt classifiers to mitigate potential misuse .
Availability and Pricing
Claude Opus 4 and Sonnet 4 are accessible through the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI. Pricing remains consistent with previous models. Opus 4 is priced at $15 per million tokens for input and $75 for output, while Sonnet 4 is available at $3 and $15, respectively .