Update — July 8, 2026: Promo extended through Sunday July 12 — extension guide
Update — July 2, 2026 (post-relaunch): First days back — developers report frequent Opus 4.8 fallbacks, rate limits reset by ClaudeDevs, and split quality vs launch week. Tips: use high effort, 10-80-10 planning, stretch the July 12 included window. Developer reaction guide →
Update — July 1, 2026: Fable 5 is live on Claude Code. Official redeploy post: global restore, new cyber classifiers — some coding/debugging routes to Opus 4.8, 50% weekly usage through July 12 then credits. Full guide → · Status hub →
Status (June 30, 2026 — Day 18): Fable 5 is not available on Claude Code for general users. June 27 Axios "within days" reporting has not produced a restore after the weekend. Anthropic's June 25 zero-traffic confirmation for Fable stands. Mythos partial restore (Annex A US critical-infra only) does not change Claude Code's consumer/developer Fable picker. Opus 4.8 remains the fallback. Live status →
Status (June 29, 2026 — Day 17): Fable 5 is not available on Claude Code for general users. Anthropic’s June 25 zero-traffic confirmation for Fable stands; no June 29 restoration post. Mythos partial restore (Annex A US critical-infra only) does not change Claude Code’s consumer/developer Fable picker. Opus 4.8 remains the fallback. Live status →
Status (June 28, 2026 — Day 16): Fable 5 is not available on Claude Code for general users. No new restoration signal from Anthropic. The day's context: Zhipu AI reports its latest model matches Claude Mythos on security bug-detection benchmarks — the same capability class cited in the export control ban. If Zhipu open-sources this model (consistent with their GLM history), internationally-accessible Mythos-parity security AI would exist outside any US export control framework. Opus 4.8 remains the Claude Code fallback. Live status →
Status (June 27, 2026 — Day 15): Fable 5 is not available on Claude Code for general users. Anthropic's official June 27 post says Mythos 5 is redeploying to US critical-infrastructure organizations first; Fable 5 general use is still being negotiated with the government. Zero Fable traffic was confirmed June 25 — that still holds. Opus 4.8 is the fallback.
June 27 official Anthropic update:
Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical infrastructure. We're restoring access for these organizations quickly, and we're continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.
Sam McAllister (Anthropic staff, June 25): "We are currently serving exactly 0 traffic to Fable 5." Amol Avasare (Head of Growth, June 25): broad access reports were categorically false. Fable 5 may still appear in the model picker (UI bug); selecting it shows "Claude Fable 5 is currently unavailable." Use Opus 4.8.
Quick Answer: Fable 5 is back on Claude Code July 1, 2026. Update to 2.1.170+, select Fable in the picker. Important: new classifiers may route some coding/debugging to Opus 4.8 with notification. 50% of weekly limits through July 7, then credits. Official terms →
What is Claude Fable 5?
Claude Fable 5 is Anthropic's 5th generation model built for your most ambitious, long-running work. It's designed to tackle days-long, complex, and asynchronous tasks that previous models couldn't sustain.
Key capabilities:
- Agents: Work for days at a time, planning across stages and checking its own work
- Coding: Handle large migrations, complex implementations, and multi-day autonomous sessions
- Vision: Understand diagrams, charts, and PDFs to check outputs against goals
- Self-Testing: Write its own tests to verify correctness
Running Fable 5 on Claude Code
According to Anthropic's official documentation, Fable 5 runs directly in Claude Code as an agent harness.
"Run Claude Fable 5 in an agent harness like Claude Code or Claude Managed Agents, and it can work for days at a time: planning across stages, delegating to sub-agents, and checking its own work."
What This Means for Developers
When you use Fable 5 in Claude Code, you get:
- Multi-Day Sessions: Fable 5 can run autonomous coding sessions that span days, not just hours
- Self-Correction: The model plans work, executes, tests itself, and iterates
- Goal Command: Use
/goalto define long-running objectives Fable 5 will work toward - Loop Engineering: Design self-correction loops where Fable 5 gets environmental feedback and improves
How to Access Fable 5
Version Requirement
Update to Claude Code v2.1.170 or later to access Fable 5.
Check your version:
claude --version
Update to the latest version:
# MacOS/Linux
curl -fsSL https://claude.ai/install.sh | bash
# Windows
irm https://claude.ai/install.ps1 | iex
# Homebrew
brew upgrade claude-code
# WinGet
winget upgrade Anthropic.ClaudeCode
If the update doesn't work, you may have multiple Claude Code installations. Clean them all up first:
# Check which version is active
which claude
claude --version
# Remove all old installations:
# 1. Remove npm installation (deprecated)
npm uninstall -g @anthropic-ai/claude-code
# 2. Remove Bun installation (if installed via Bun)
bun remove -g @anthropic-ai/claude-code
# 3. Remove Homebrew npm symlinks (check /opt/homebrew/bin/)
/opt/homebrew/bin/npm uninstall -g @anthropic-ai/claude-code
# 4. Remove Homebrew cask (if installed)
brew uninstall --cask claude-code
# Verify all removed
which claude
# Should show: /Users/yourusername/.local/bin/claude (or nothing)
# Then reinstall using recommended method:
# MacOS/Linux
curl -fsSL https://claude.ai/install.sh | bash
# Windows
irm https://claude.ai/install.ps1 | iex
# Verify correct version installed
claude --version
# Should show: 2.1.170 (Claude Code)
Note: You may need to close and reopen your terminal after uninstalling to refresh the PATH.
For Organizations
Claude Fable 5 is available on the consumption-based Enterprise plan.
For Developers
Access Fable 5 through:
- Claude Platform API (native)
- AWS (Amazon Web Services)
- Google Cloud (Vertex AI)
- Microsoft Foundry
Model identifier: claude-fable-5
Pricing
- Input tokens: $10 per million tokens
- Output tokens: $50 per million tokens
- Prompt caching: 90% discount on cached input tokens
- US-only inference: 1.1x pricing multiplier
What Makes Fable 5 Special for Coding?
Fable 5 is the most capable model for ambitious coding projects:
1. Large Migrations
Refactor entire codebases, migrate frameworks, or update dependencies across hundreds of files.
2. Complex Implementations
Build features that require architectural decisions, multiple file changes, and thorough testing.
3. Self-Verification
Fable 5 writes tests to check its own work, ensuring code correctness before you review.
4. Vision Integration
Uses vision to check code outputs against design mockups or architecture diagrams.
Using Fable 5 with Claude Code Features
Goal Command
The /goal command enables Fable 5 to work autonomously toward objectives:
/goal "Migrate entire codebase from JavaScript to TypeScript"
Fable 5 will:
- Analyze the current codebase
- Plan the migration strategy
- Convert files incrementally
- Run tests after each change
- Fix type errors
- Verify 100% TypeScript coverage
Self-Correction Loops
Learn how to design effective loops where Fable 5 gets feedback and improves:
- Uses verifier sub-agents (better than self-critique)
- Follows fail → investigate → verify → distill → consult progression
- Achieves 6x improvements over earlier models on complex tasks
Important: Safeguards & Data Retention
Automatic Safeguards
Fable 5 includes robust safeguards for cybersecurity and biology. Queries flagged in these domains are automatically routed to Opus 4.8 (you won't be charged Fable prices for rerouted requests).
Data Retention
Using Fable 5 requires 30-day data retention for safety monitoring. Learn more about data retention policies.
Benchmarks: Fable 5 vs Other Models
Claude Fable 5 is state-of-the-art at coding, knowledge work, vision, and computer use.
According to Anthropic's benchmarks, Fable 5 outperforms other leading models on:
- SWE-Bench (coding challenges)
- Agent tasks (multi-step workflows)
- Vision understanding (diagrams, charts, PDFs)
- Long-horizon tasks (days-long projects)
Should You Use Fable 5 on Claude Code?
Use Fable 5 when:
- You have a complex, multi-day coding project
- You need to migrate or refactor large codebases
- You want the model to work autonomously with minimal supervision
- The task requires structural changes, not just simple edits
Skip Fable 5 for:
- Quick one-off edits or bug fixes
- Simple tasks that Claude Opus or Sonnet can handle
- Budget-sensitive projects (Fable 5 is premium-priced)
Getting Started with Fable 5 on Claude Code
- Update: Upgrade to Claude Code v2.1.170 or later (run
claude --versionto check)- If update fails, uninstall (
npm uninstall -g @anthropic-ai/claude-codeorbrew uninstall claude-code) and reinstall
- If update fails, uninstall (
- Access: Ensure you have API access or Enterprise plan subscription
- Model Selection: Use
claude-fable-5as your model identifier - Define Goals: Use
/goalfor long-running autonomous work - Monitor Progress: Fable 5 will update you as it works through tasks
- Review Output: Check completed work rather than supervising every step
Troubleshooting: Multiple Claude Code Installations
Problem: Wrong Version After Update
If claude --version shows an old version (like 2.1.23) even after installing 2.1.170, you likely have multiple Claude Code installations in your PATH.
How to Fix
Step 1: Identify all installations
which claude
# Example output: /Users/yash/.bun/bin/claude
Step 2: Remove ALL old installations
You may have Claude Code installed through:
- npm (deprecated method)
- Bun (package manager)
- Homebrew (symlinks to npm)
- Multiple npm locations (nvm, system npm, Homebrew npm)
Remove them all:
# npm (check all locations)
npm uninstall -g @anthropic-ai/claude-code
/opt/homebrew/bin/npm uninstall -g @anthropic-ai/claude-code
# Bun
bun remove -g @anthropic-ai/claude-code
# Homebrew cask
brew uninstall --cask claude-code
# Verify which is still active
which claude
Step 3: Clean install
# MacOS/Linux (recommended method)
curl -fsSL https://claude.ai/install.sh | bash
# This installs to: ~/.local/bin/claude
# Restart your terminal
# Verify correct version
claude --version
# Should show: 2.1.170 (Claude Code)
Expected Installation Location
After a clean install, which claude should show:
- MacOS/Linux:
/Users/yourusername/.local/bin/claude - Windows:
C:\Users\YourName\.local\bin\claude.exe
If it shows anything else (like /opt/homebrew/bin/claude or /.bun/bin/claude), you still have an old installation taking precedence.
Related Reading
- Fable 5 Biology Safeguards Update: 85% Fewer False Fallbacks (August 2026)
- Claude Fable 5 and Mythos 5 Launch
- Designing Loops with Fable 5: Self-Correction and Memory Guide
- Claude Code Goal Command for Long-Running Agents
- Loop Engineering: Coding Agent Loops That Run While You Sleep
- Claude Fable 5 Creates Minecraft Clones from Simple Prompts
Bottom Line: Claude Fable 5 represents a major leap in autonomous coding capabilities. Use it for your most ambitious projects where you need multi-day autonomy, self-correction, and frontier intelligence.
Update (June 25, 2026): Anthropic staff debunked viral return rumors — zero Fable traffic on Claude Code. Model picker UI bug only; selecting Fable 5 shows unavailable. Opus 4.8 fallback. Live status →
