Anthropic hasn’t been chasing headline model releases lately so much as quietly rebuilding what sits around the model — how much it thinks, how it talks, and how it fits into a normal workday. New Claude features in 2026 lean heavily into that direction: a smarter Opus release, real control over how hard Claude thinks about a task, a more natural voice mode, and a growing push to get Claude working across your whole day rather than one chat window at a time.
Claude Opus 4.8
Released May 28, 2026, Opus 4.8 improved on coding, agentic tasks, and long-running professional work at the same price as its predecessor. Alongside the model itself, Anthropic shipped effort control — a dropdown next to the model selector on claude.ai that lets you directly set how much reasoning effort Claude puts into a task, from a fast response to maximum depth. Opus 4.8 defaults to a “high” setting, described as the balance point between quality and response time.
Claude Reflect: A Usage Dashboard
Introduced July 9, 2026, Reflect is a built-in dashboard that shows how you actually use Claude — the topics you discuss most, your usage patterns over time, and the kinds of tasks you tend to hand off to it. It’s currently in beta for Free, Pro, and Max users who have memory turned on, and it’s a genuinely useful way to notice your own habits rather than a productivity tracker in the traditional sense.
Upgraded Voice Mode
As of July 23, 2026, Claude’s voice mode lets you choose between Opus, Sonnet, and Haiku for spoken conversations, rather than being locked to the lightweight Haiku model it launched with. It now defaults to the fastest version of whichever model you were last using in text chat, and handles longer, more complex spoken conversations noticeably better than the original release.
Cowork Expands to Mobile and Web
Cowork — Anthropic’s hand-off-a-task interface that works across your files, calendar, and email — is expanding beyond desktop to mobile and web, letting sessions and files follow you across devices. The rollout includes background work, scheduled tasks, shared chats and projects, and mobile approvals, starting with Max subscribers before reaching other plans.
Skills 2.0 and a Bigger Connector Directory
- Skills as workflow packages. Skills have moved beyond saved prompts into full workflow packages that bundle instructions, scripts, templates, and reference materials into one unit Claude can execute on demand — capable of running scripts and producing real deliverables, not just text.
- 950+ MCP connectors. Claude’s connector directory has grown past 950 listed MCP servers, alongside an updated MCP specification that makes it easier for companies to build and scale their own integrations.
Frequently Asked Questions
What does effort control actually do in Claude?
It lets you set how much reasoning Claude applies to a task directly from a dropdown next to the model selector, trading off between faster responses and deeper, more thorough answers.
Is Claude Reflect available to free users?
Yes, it’s in beta for Free, Pro, and Max users, as long as memory is turned on for the account.
What changed in Claude’s voice mode?
You can now choose between Opus, Sonnet, and Haiku for voice conversations, rather than being limited to the original Haiku-only version, and it automatically defaults to your most recently used model.
Where can I learn the basics of using Claude?
Our guide on how to use Claude covers setup and everyday use, and how to get Claude Pro free covers legitimate ways to try the paid tier.
Final Thoughts
Claude’s 2026 updates read less like a single big leap and more like Anthropic sanding down the rough edges — giving people direct control over reasoning depth, a more natural voice experience, and a workspace that follows them off the desktop. If you’ve mostly used Claude for one-off chat questions, effort control and the Cowork expansion are worth trying first. For how the competition is moving, see our roundups on new ChatGPT features and new Grok features.
