Claude Opus 5 and Sonnet 5 launched (July 24 and June 30, 2026). Opus 5 delivers Anthropic’s highest reasoning performance for complex analysis, while Sonnet 5 balances speed and capability for everyday professional work. Source: Anthropic release notes, July 2026.
Claude Cowork expands to web and mobile (July 7, 2026). The agentic workspace that lets Claude use your screen, apps, and files is now available on claude.ai and the iOS/Android apps, not just the desktop client. Source: Anthropic release notes, July 7, 2026.
Upgraded voice mode with Opus and Sonnet (July 24, 2026). Voice mode now supports Opus 5 and Sonnet 5, connects to tools like Gmail and Slack, and works in more languages. Source: Anthropic release notes, July 24, 2026.
Record a Skill feature learns from your screen. A new capability lets you record a workflow on your screen and teach Claude to repeat it as a reusable skill. Source: Anthropic YouTube demo, July 2026.
Self-serve HIPAA configuration (July 14, 2026). Organizations on the Enterprise plan can now enable HIPAA compliance and sign a Business Associate Agreement (BAA) directly from the admin console without contacting sales. Source: Anthropic release notes, July 14, 2026.
Memory, Monthly Recap, and Reflection tools (July 9, 2026). Claude can now remember preferences across sessions, generate a monthly summary of your work, and guide structured reflection on projects. Source: Anthropic release notes, July 9, 2026.
Microsoft 365 Connector and admin controls (July 7, 2026). Write tools for Word, Excel, and PowerPoint are now available through the Microsoft 365 connector, alongside new admin analytics, spend controls, and model-level entitlements. Source: Anthropic release notes, July 7, 2026.
Claude AI Updates Introduce Opus 5 and Sonnet 5 Models
Claude Opus 5 is Anthropic’s most capable model for deep reasoning, multi-step analysis, and complex code generation. Claude Sonnet 5 is the optimized workhorse for professional tasks that need a balance of speed, quality, and cost. Both models are available now on Claude Pro, Max, Team, and Enterprise plans.
The Opus 5 launch on July 24, 2026, followed the Sonnet 5 release on June 30, 2026. Opus 5 excels when you need to analyze dense legal contracts, debug multi-file codebases, or synthesize research across dozens of documents. Sonnet 5 handles the majority of daily professional work—drafting reports, answering customer inquiries, and processing structured data—with lower latency and lower cost per query.
A key architectural difference is the context window. Opus 5 supports up to 1 million tokens, enough to ingest entire code repositories or book-length manuscripts in a single session. Sonnet 5 operates with a 200,000-token context window, which is sufficient for most document-length tasks. The choice between them is a trade-off: Opus 5 for maximum depth on the hardest problems, Sonnet 5 for responsive throughput on high-volume work.
Both models are accessible through the Claude web app, mobile apps, API, and cloud platforms including Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Anthropic’s release notes confirm that Opus 5 and Sonnet 5 are the default models for new conversations on supported plans.
Claude Model Comparison: July 2026
| Model | Type | Context Window | Coding Performance | Recommended Use Case |
|---|
| Claude Opus 5 | Maximum reasoning | 1,000,000 tokens | Highest accuracy on complex multi-file tasks | Deep research, legal analysis, complex codebase work |
| Claude Sonnet 5 | Balanced workhorse | 200,000 tokens | Strong performance with lower latency | Daily professional writing, analysis, and coding |
| Claude Opus 4.6 (previous) | High reasoning | 1,000,000 tokens | Strong but slower than Opus 5 | Legacy complex tasks; superseded by Opus 5 |
If your work involves occasional deep dives but mostly routine tasks, start with Sonnet 5 and switch to Opus 5 only when you hit a problem that requires extended reasoning. Teams on Enterprise plans can set model-level entitlements so users default to the right model for their role.
Claude AI Updates Bring Cowork to Web and Mobile
Claude Cowork is an agentic workspace where Claude can see your screen, use your applications, read and edit files, and execute multi-step workflows autonomously. As of July 7, 2026, Cowork is available on the web at claude.ai and in the Claude mobile apps for iOS and Android.
Previously, Cowork was limited to the Claude desktop application. The expansion means you can start a Cowork session on your phone, hand off a task, and pick up the results on your laptop. Sessions and files sync across devices, so a research project begun on mobile is fully available on the web.
Cowork is designed for tasks that require sustained attention across multiple tools. For example, you can ask Claude to research a competitor, draft a memo in Google Docs, pull data from a spreadsheet, and email the result to your team—all within one Cowork session. The agent can navigate your screen, click buttons, and type into applications with your permission.
A practical constraint: Cowork works best with clearly defined, sequential tasks. Ambiguous instructions or tasks that require real-time human judgment mid-stream can cause the agent to stall. Start with a small, well-scoped task before handing off a complex workflow.
Voice mode in Claude now supports Opus 5 and Sonnet 5 models, connects to integrated tools, and operates in more languages. The upgrade shipped on July 24, 2026, alongside the Opus 5 launch.
The most significant change is tool connectivity. In previous versions, voice mode was a conversational interface only. Now, you can speak a command like “Check my Gmail for the latest invoice from the supplier and summarize it,” and Claude will access Gmail through the connector, retrieve the email, and read the summary back to you. The same works with Slack, Google Calendar, and other connected tools.
Multilingual support has also expanded. Anthropic’s release notes confirm that voice mode now works in “many more languages,” though the exact list is not published. If you work across languages, test voice mode with your specific language pair to confirm accuracy before relying on it for critical communication.
Voice mode is available in the Claude mobile app. Download the latest version from the iOS App Store or Google Play Store, then tap the microphone icon to start a voice session. You can switch between Opus 5 and Sonnet 5 during a voice conversation, which is useful when a question starts simple but deepens into something that needs heavier reasoning.
Claude AI Updates Add Record a Skill for Workflow Automation
Record a Skill is a feature that lets you record a sequence of actions on your screen and save it as a reusable skill that Claude can execute on demand. Anthropic demonstrated the feature in a July 2026 YouTube video.
The workflow is straightforward: you start a recording, perform the task once—such as pulling data from a CRM, formatting it in a spreadsheet, and emailing a report—and then save the recording as a named skill. Later, you or a teammate can trigger that skill with a natural-language command, and Claude will repeat the steps.
This feature is most valuable for repetitive multi-step processes that follow a consistent pattern. Examples include weekly report generation, invoice processing, data entry from emails into a database, and standardized customer onboarding checklists. It is less suited to tasks where each instance requires significant judgment or variation.
Because Record a Skill captures screen interactions, it works with any application visible on your screen, not just those with official Claude connectors. This broadens the range of tools Claude can automate, but it also means the feature is sensitive to changes in your interface layout. If a button moves or a menu changes, the recorded skill may break and need re-recording.
Claude AI Updates Enable Self Serve HIPAA Configuration
Self-serve HIPAA configuration allows Enterprise plan administrators to enable HIPAA compliance and execute a Business Associate Agreement (BAA) directly from the Claude admin console, without contacting Anthropic’s sales team. This shipped on July 14, 2026.
Previously, enabling HIPAA compliance required a manual sales process that could take days or weeks. The self-serve option reduces that to minutes. Administrators navigate to the compliance settings in the admin console, toggle HIPAA mode, review and sign the BAA electronically, and the organization’s Claude workspace is immediately configured for protected health information (PHI).
This matters for healthcare providers, insurers, and health-tech companies that need to use Claude for tasks involving patient data, clinical notes, or medical research. Without HIPAA mode, Claude’s standard data handling does not meet the regulatory requirements for PHI under U.S. law.
Important limitations: HIPAA mode is available only on the Enterprise plan, not on Pro, Max, or Team plans. It applies to the Claude workspace and API, but you must still configure your own systems and workflows to maintain end-to-end compliance. Anthropic’s BAA covers its processing of PHI, but your organization remains responsible for how you collect, store, and use that data before it reaches Claude.
Claude AI Updates Introduce Memory Monthly Recap and Reflection
Memory lets Claude remember facts and preferences across conversations. Monthly Recap generates a summary of your Claude usage and projects. Reflection tools guide structured review of your work. All three shipped on July 9, 2026.
Memory works by storing information you explicitly tell Claude to remember, such as your name, role, preferred writing style, or ongoing project context. Claude surfaces these memories in relevant conversations, reducing the need to re-explain context. You can view, edit, and delete memories at any time from the settings panel.
Monthly Recap provides a digest of your Claude activity: which projects you worked on, which models you used most, and how many conversations you had. It is designed to help you reflect on your AI usage patterns and identify areas where Claude is adding the most value—or where you might be underusing available features.
The Reflection tools are more structured. They prompt you with questions about recent work, help you identify patterns, and suggest next steps. This is not passive summarization; it is an interactive process where Claude asks you to evaluate your own output and decisions. Teams on Enterprise plans can use Reflection in shared project spaces to conduct lightweight retrospectives.
Claude AI Updates Deliver Microsoft 365 Connector and Admin Controls
Microsoft 365 connector now includes write tools for Word, Excel, and PowerPoint, allowing Claude to create and edit documents directly in your Microsoft 365 environment. New admin controls give Enterprise administrators visibility into usage, spend, and model access. Both shipped on July 7, 2026.
The write tools are the key addition. Previously, the Microsoft 365 connector could read from Word, Excel, and PowerPoint. Now, Claude can also create new documents, edit existing ones, and save changes back to your OneDrive or SharePoint. This turns Claude from a read-only assistant into a full document collaborator within the Microsoft ecosystem.
For administrators, the July release adds several controls. Usage analytics show adoption rates, active users, and model usage patterns across the organization. Spend controls let you set budgets and receive alerts when usage approaches a threshold. Model-level entitlements allow you to specify which users or groups can access Opus 5 versus Sonnet 5, preventing unexpected costs from high-compute model usage.
These controls are available in the admin console for Team and Enterprise plans. If your organization uses Microsoft Entra for identity management, you can integrate Claude’s admin controls with your existing directory for unified user provisioning and access management.
