Claude Latest News 2026 | What Changed with Fable 5, Design, and Code
Anthropic's Claude has seen a rapid series of changes in 2026: the launch of a new flagship model, the release of a design generation tool, and expanded automation capabilities in Claude Code. This article consolidates the major updates as of June 2026 and breaks down which changes matter for which users.
There are three main topics for June 2026. (1) Claude Fable 5 became generally available on June 9, becoming the current top-performing model with Adaptive Thinking always on and a 1M context window. (2) Claude Design launched in April, enabling users to generate slides and UI prototypes simply by providing text or documents as input. (3) Claude Code has introduced a series of features that enhance autonomous execution — including Dynamic workflows, Auto mode, and /goal — creating an environment where long-running batch tasks can proceed without user intervention. Note that Sonnet 4 and Opus 4 are scheduled for deprecation on June 15, so API users need to migrate promptly.
目次 (13)
- Claude Fable 5 — The Top-Tier Model as of June 2026
- Claude Design — Generate Design Artifacts from Prompts
- What You Can Do with Claude Design
- Supported Plans and Technical Specifications
- Claude Opus 4.8 Elevated to Default Model
- Major Claude Code Updates (April–June 2026)
- Dynamic Workflows — Automation for Large-Scale Parallel Processing
- Auto Mode Now Available for Pro Plan
- New Commands: /usage, /goal, /code-review
- Agent View — Session List Management
- /ultrareview — Parallel Cloud Code Review
- Checking Deprecated Models
- Summary: How to Make the Most of the New Features
Claude Fable 5 — The Top-Tier Model as of June 2026
On June 9, 2026, Anthropic made Claude Fable 5 generally available (source: Anthropic Official Model List). It is accessible across the API, AWS Bedrock, Vertex AI, and Microsoft Foundry.
| Item | Specification |
|---|---|
| API ID | claude-fable-5 |
| Context Window | 1,000,000 tokens |
| Max Output Tokens | 128,000 |
| Input Pricing | $10 / 1MTok |
| Output Pricing | $50 / 1MTok |
| Knowledge Cutoff | January 2026 |
| Thinking Feature | Adaptive Thinking (always on) |
With the previous generation, users had to manually enable Extended Thinking, but Fable 5's Adaptive Thinking is always on. The model autonomously judges task difficulty and adjusts reasoning depth — conserving tokens on simple tasks while allocating sufficient reasoning for complex ones.
Additionally, an invite-only research preview called Claude Mythos 5 (API ID: claude-mythos-5) is available in parallel, accessible only to approved customers through Project Glasswing. Its specifications are largely on par with Fable 5, but it includes early experimental features.
Claude Design — Generate Design Artifacts from Prompts
On April 17, 2026, Anthropic Labs announced Claude Design (source: Anthropic News). It is a new tool that takes text, images, and documents as input and generates visual outputs such as designs, prototypes, slides, and one-pagers.
What You Can Do with Claude Design
- Generate slides and design assets from text prompts alone
- Fine-tune results after generation using inline comments or direct editing
- Adjust styles and colors with custom sliders
- Automatically apply your team's design system (for Enterprise)
- Export to Canva, PDF, PPTX, or standalone HTML
Supported Plans and Technical Specifications
Rolling out progressively to Claude Pro, Max, Team, and Enterprise subscribers. Available within the subscription at no additional cost, with add-on options provided when usage limits are exceeded. Internally powered by the vision capabilities of Claude Opus 4.7.
The biggest appeal is the ability to create materials without any design skills — envisioned for use by marketers and salespeople as an alternative to Figma or PowerPoint.
Claude Opus 4.8 Elevated to Default Model
On May 29, 2026 (Claude Code Week 22), Claude Opus 4.8 was elevated to the default model for Max, Team Premium, Enterprise, and Anthropic API accounts (source: Claude Code What's New).
Key specifications are as follows:
- Context Window: 1,000,000 tokens
- API Pricing: $5 / $25 (input/output per MTok)
- Default
effortlevel:high - Option for hardest tasks:
/effort xhigh - Fast mode (research preview): high-speed operation at $10 / $50 per MTok
Compared to Fable 5, pricing is roughly half, making Opus 4.8 a cost-effective choice for everyday development work. Fable 5 is best reserved for scenarios requiring long-context processing or advanced reasoning.
Major Claude Code Updates (April–June 2026)
Dynamic Workflows — Automation for Large-Scale Parallel Processing
In a late May 2026 update, Dynamic workflows were added, enabling scripts to orchestrate dozens to hundreds of subtasks. Useful for large-scale code migrations, parallel test runs, and tasks spanning multiple repositories. Simply write a workflow script and run it in the background — Claude Code manages progress and processes tasks automatically.
Auto Mode Now Available for Pro Plan
In updates from May 18–22, 2026, Auto mode was opened to Claude Pro plan users. By replacing permission confirmation prompts with background safety checks, it minimizes the need for manual user approval while continuing to work autonomously. It automatically selects between Sonnet 4.6 and Opus models based on the situation.
New Commands: /usage, /goal, /code-review
Several new commands were also introduced around the same time:
/usage: Displays a breakdown of token usage by skill, plugin, and MCP server, making cost analysis easier/goal: Set a goal condition, and Claude will continue working autonomously across multiple turns until that condition is met/code-review: A command for reporting correctness bugs in code, supporting multiple investigation depths fromlowtoultra
Agent View — Session List Management
The claude agents command was added in updates from May 11–15, 2026. It lets you view all running, pending, and completed Claude Code sessions in a single screen, making it much easier to monitor when running multiple instances in parallel.
/ultrareview — Parallel Cloud Code Review
Released as a public research preview in late April 2026. Multiple review-specialized instances run in parallel in the cloud to report bugs, performance issues, and security concerns together. Use it by running /code-review ultra or calling claude ultrareview from a CI script.
Checking Deprecated Models
Anthropic plans to deprecate the following models:
| Model | Deprecation Date |
|---|---|
| Claude Sonnet 4 | June 15, 2026 |
| Claude Opus 4 | June 15, 2026 |
| Claude Opus 4.1 | August 5, 2026 |
Users who have Sonnet 4 or Opus 4 hard-coded in their API calls need to switch to claude-opus-4-8 or claude-fable-5 by June 15. After deprecation, API calls will return errors, so verify your migration now to avoid any gaps.
Summary: How to Make the Most of the New Features
Here's a prioritized breakdown for getting the most out of Claude as of June 2026:
- API users (urgent): If you're using Sonnet 4 or Opus 4, switch to
claude-opus-4-8orclaude-fable-5before the June 15 deprecation - Claude Code users: Use
/usageto visualize costs, and expand autonomous execution with Auto mode and/goal - Design and document creation work: On Claude Pro or higher plans, wait for the Claude Design rollout and start by testing small-scale document generation
- Advanced reasoning and long-context processing: Try Claude Fable 5 and evaluate how much Adaptive Thinking's auto-adjustment delivers in real-world use cases
The pace of Claude's evolution has accelerated into 2026, and keeping up with new features takes effort. A practical approach is to regularly reference summary articles like this one and selectively adopt only the changes most relevant to your own use cases.
