Claude Fable 5 Arrives in Claude Code: Anthropic's First Mythos-Class Model for General Use

A New Model Tier: Claude Fable 5

With the release of v2.1.170, Claude Code users gain access to Claude Fable 5: Anthropic's most capable model ever made generally available. Fable 5 is drawn from the Mythos model class, a tier that sits above the Opus family in raw capability. While Claude Mythos 5 itself remains in limited access through Project Glasswing, Fable 5 delivers the same underlying capabilities with Anthropic's safety classifiers applied, making it safe for broad developer use.

According to Anthropic, Fable 5 achieves state-of-the-art results on nearly every major benchmark tested, with particular strengths in the areas most relevant to Claude Code workflows: software engineering, vision-based tasks, and long-context reasoning.

What Fable 5 Can Do That Previous Models Could Not

Extended Autonomous Engineering

Fable 5 is designed to work autonomously for longer stretches than any previous Claude model. Stripe reported that Fable 5 completed a codebase-wide migration across a 50-million-line Ruby codebase in a single day: a task that would otherwise have taken a whole engineering team over two months. The model maintains coherent focus across millions of tokens, making it significantly more reliable for large-scale refactors, multi-file edits, and complex agent workflows.

Vision Capabilities

Fable 5 sets a new bar for vision-based tasks. It can extract precise data from detailed scientific figures, rebuild web application source code from screenshots alone, and handle multimodal inputs at a level no previous Claude model achieved. For Claude Code users working on design-to-code workflows or visual debugging tasks, this is a meaningful upgrade.

Long-Context and Memory Performance

On memory-dependent tasks, Fable 5 improved threefold compared to earlier models. It can leverage persistent memory across sessions more effectively, making it better suited to long-running projects where context continuity matters.

Pricing and Plan Availability

Fable 5 is available immediately via the Claude API. Through June 22, 2026, the model is included at no additional cost on Pro, Max, Team, and seat-based Enterprise plans. Starting June 23, usage credits will be required for Fable 5 access on subscription plans.

The API pricing is $10 per million input tokens and $50 per million output tokens: less than half the price of Claude Mythos Preview, though still twice the cost of Claude Opus 4.8.

One consideration worth noting: Fable 5 falls back to Claude Opus 4.8 in a small number of high-risk categories (cybersecurity, biology, chemistry, and similar domains). Early data from Anthropic shows this fallback occurring in fewer than 5% of sessions.

Bug Fix: VS Code Terminal Transcript Saving

This release also resolves a persistent bug where Claude Code sessions launched from the VS Code integrated terminal, or from any shell that inherited Claude Code's environment variables, failed to save transcripts. Affected sessions would not appear in --resume output. The fix restores normal transcript saving behavior for all launch contexts.