Anthropic has launched its Claude 4 models—Opus 4 and Sonnet 4—on iOS, bringing high-performance AI to mobile users. This rollout expands access to state-of-the-art coding, reasoning, and autonomous agent capabilities. You can now use Claude’s latest models directly from your iPhone, with the same extended reasoning and tool-use features available on desktop.
Claude Opus 4 is Anthropic’s most powerful model to date. It leads benchmarks with a 72.5% score on SWE-bench and 43.2% on Terminal-bench, outperforming all previous Sonnet models and competing systems. The model sustains high performance on long-running, complex tasks. Anthropic says Claude 4 worked autonomously for seven hours in customer tests, showing consistent performance without losing context or making critical errors.
Opus 4 and Sonnet 4: High-Performance Models
Claude Opus 4 powers advanced agent products and is designed to handle coding tasks across large, multi-file projects. It can reason across thousands of steps and store long-term memory when granted file access. Rakuten tested it on a demanding open-source refactor that ran independently for seven hours, confirming its ability to maintain focus and accuracy over extended periods.
Claude Sonnet 4, also available to free-tier users, improves on version 3.7 with a top score of 72.7% on SWE-bench. It offers faster response times and better steerability. GitHub plans to use Sonnet 4 in the next version of GitHub Copilot, citing improvements in agent performance, problem-solving, and code quality. Companies like Manus and iGent also reported gains in instruction following, app development, and error reduction.
Claude Code Expands Developer Tools
Claude Code, now generally available, integrates with VS Code and JetBrains. You can run Claude in your IDE’s terminal, see inline edits, and automate tasks using GitHub Actions. The new Claude Code SDK lets you build custom agents and applications with the same core engine. In beta, developers can tag Claude Code on GitHub pull requests to address reviewer feedback or CI errors.
Both models now support extended thinking, parallel tool use, and local memory files. When given access, Claude can create “memory files” that help maintain continuity across sessions—for example, logging key steps while playing Pokémon Red.
As reported by Anthropic, these updates aim to turn Claude into a true collaborator that thinks longer, reasons deeper, and stays on track. With Claude 4 now on iOS, you can use these capabilities anytime, anywhere—whether you’re writing code, reviewing documents, or deploying AI agents on the go.