Claude Fable 5 launches, Anthropic files for IPO, Partner Hub goes live

Anthropic released Claude Fable 5 on June 9, the first publicly available model in the Mythos class — the most capable tier Anthropic has ever shipped. Fable 5 targets days-long, complex, autonomous tasks and posts 80.3% on SWE-Bench Pro (vs 69.2% for Opus 4.8) and is the first model to exceed 90% on Hex's analytical benchmark. In high-risk areas like cybersecurity, biology, and chemistry, the model automatically falls back to Claude Opus 4.8 — a hard safety guardrail that triggers in under 5% of sessions. Priced at $10 per million input tokens and $50 per million output tokens (90% discount with prompt caching — exactly double Opus pricing). Available immediately on the Claude API (model ID: claude-fable-5), Claude Platform, Claude Code, consumption-based Enterprise plans, AWS, Google Cloud, and Microsoft Foundry. Free on Pro, Max, Team, and seat-based Enterprise plans through June 22; usage credits required from June 23.

Alongside Fable 5, Anthropic simultaneously released Claude Mythos 5 to organizations already approved under Project Glasswing. Mythos 5 shares the same underlying training as Fable 5 but with certain protections removed, reserved for vetted cybersecurity professionals, researchers, and Glasswing partners including AWS, Microsoft, Apple, and CrowdStrike. Anthropic says Mythos 5 delivers state-of-the-art performance on coding, reasoning, vision, scientific research, and long-context tasks without the hard-blocked domains that constrain Fable.

Anthropic filed a confidential draft registration statement on Form S-1 with the U.S. Securities and Exchange Commission on June 1, setting up a potential IPO. The filing comes less than a week after the company closed a $65 billion Series H round — co-led by Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital, Capital Group, Coatue, and D1 Capital Partners — which pushed its valuation to $965 billion, overtaking OpenAI's $852B. Revenue run rate has reached $47 billion, up from $10B in annual revenue last year. No share count or price range has been set. Anthropic is expected to join SpaceX and OpenAI as one of three potential trillion-dollar listings of 2026.

Anthropic launched the Services Track and Claude Partner Hub — a formal three-tier certification system for consulting and implementation partners. More than 40,000 firms have applied since the Claude Partner Network launched in March with a $100M investment; over 10,000 consultants have already earned certification. The Hub gives partners daily-updated dashboards showing their standing against published tier requirements, while customers get a public directory with each firm's certifications, production deployments, and references. A new MCP connector links the Hub directly with Claude so partners can ask about certification status and advancement without leaving the assistant. Tier promotions run on Jan. 1 and July 1, with an additional review on Oct. 1, 2026.

Claude Code's latest release adds --safe-mode (and CLAUDE_CODE_SAFE_MODE env var) to start sessions with all customizations disabled for clean troubleshooting. A new /cd command moves a session to a new working directory without breaking the prompt cache mid-session. Prefix any shell command with ! in claude agents to run it as a background job you can attach and detach from — also available as claude --bg --exec. Plugins in .claude/skills directories now load automatically without marketplace registration, and a new /reload-skills command re-scans skill directories without restarting the session.

Claude Code now supports fallback models when the primary model is unavailable — particularly useful for teams running Fable 5 with Opus 4.8 as the fallback. Broader deny-rule glob support lands for more precise policy enforcement. New requiredMinimumVersion and requiredMaximumVersion managed settings let enterprises lock Claude Code to a specific version range — the CLI refuses to start outside that range and directs users to an approved version. Also fixed: cursor sticking at end of first line in multiline claude agents prompts, and blank lines between agent rows on terminals without Unicode support.

Claude Security is a new first-party plugin for Claude Code that runs a fast pattern check on each edit, a model review at the end of each turn, and a deeper agentic review on commit or push. Project-level rules go in .claude/claude-security-guidance.md. Install via /plugin install security-guidance@claude-plugins-official from the official Anthropic marketplace. DoorDash's CSO called it out specifically: 'It surfaces deep vulnerabilities accurately, and pipes findings right into our workflows so engineers can act on them in context.'

Anthropic released a new Swift package that lets Apple developers use Apple's Foundation Models framework to call Claude for complex workflows. Add the package to your Xcode project, sign in with an Anthropic API key, and pass typed outputs from Apple's on-device model into a Claude request — the package handles streaming, tool calls, and structured responses back into SwiftUI. Works on iOS 27, iPadOS 27, macOS 27, visionOS 27, and watchOS 27.

Dynamic Workflows are now available for early access users on Claude Platform, handling tasks too large for a single agent pass: codebase-wide bug hunts, framework migrations spanning thousands of files, security audits across entire repos, and profiler-guided optimization. Claude searches in parallel, runs independent verification on every finding, and surfaces a consolidated report of real issues. Early access users and Anthropic internal teams have been running these workflows across codebases, with support for attaching and detaching mid-workflow.

Enterprise plan admins can now configure custom roles that grant billing or privacy admin access without granting full Owner access. This extends the existing custom roles framework — previously custom roles could only control what members could do with Claude, not platform administration. Useful for compliance teams and IT admins who need billing visibility or privacy controls without full org ownership.

The Code with Claude roadshow concluded its Tokyo leg on June 5–6, with Day 1 keynotes streamed live. Sessions covered agentic development with teams from Asana, Cursor, GitHub, Replit, and Vercel. The event followed the successful London edition in May. No further dates announced yet.