The news | One underlying model, two access arrangements

On November 9, 2026, Anthropic released Claude Fable 5.2 and Claude Mythos 5.2. Claude models understand material, generate content, and handle tasks such as coding. These two names share an underlying model but have different access and safeguard arrangements. Fable reaches a broader audience, while Mythos requires specific verification programs for life sciences or cybersecurity. Official announcement

Anthropic reports improvements in coding, reasoning, and vision, output up to 30% faster than 5.1, and fewer unnecessary safety blocks. These are vendor-reported changes; use your own tasks to judge whether an upgrade helps.

Why it helps | Check the task and entry point before switching

Suppose your regular model repeatedly struggles with a code change that must fix a known bug while preserving existing behavior. Give an accessible Fable 5.2 instance the known inputs, expected results, and boundaries, then check whether the change meets them. Published benchmark performance can explain positioning but cannot validate this particular code.

Two names do not mean an ordinary account can switch freely between them. Without eligibility for a Mythos verification program, use the Fable entry actually available to you. The official recommendation still keeps Opus 5.5 as a daily model and upgrades for harder work.

How to use it | Try a task with a checkable result

This example uses Claude Code. It requires a paid account with Fable 5.2 access and version 2.1.321 or later. Team and Enterprise also depend on organization model settings; Free is ineligible. Plans may use included allowances or metered credits, so check usage and billing first. Launch promotions apply only during their stated periods and to eligible plans. Access conditions

  1. Start a session and choose the model. Open a session in Claude Desktop’s Code tab, or run claude in the project terminal. Enter /model and select Fable 5.2; use /status to confirm the current model.
  2. Give it a verifiable change. Supply the relevant function, failing input, and known correct output. State which earlier behaviors must remain, then send the request.
  3. Check the artifact. Inspect changed files, run existing validation, and retry the failing input. Consider the explanation of key changes when deciding whether behavior meets the request.
  4. Decide how to use it afterward. If it improves this kind of difficult work, retain that division of tasks. Choose daily models by cost and results; Mythos still has separate verification requirements. Switching models in Claude Code