Anthropic’s Claude Code changelog lists version 2.1.321, dated 9 November 2026, with one headline change: Claude Fable 5.2 (claude-fable-5-2) is now the default Fable model. Saved Fable 5.1 model choices move over to the generic fable setting.
What the changelog states
- A 1M-token context window, enough to hold a large codebase in a single working session.
- Pricing of $8 per million input tokens and $40 per million output tokens, with cached input reads at $0.20 per million tokens.
- A new
--settings <file>option forclaude remote-control, so a team can apply one settings file to every remote session it starts. - Mods’ own tool calls no longer raise permission prompts, although deny rules and organisation policy still apply.
What it means
Two numbers matter more than the model name. The first is the cache-read price: at $0.20 per million tokens it is a fraction of the headline input price, which is what makes long, repeated sessions over the same codebase affordable. If you commission AI-assisted development, ask your supplier whether their workflow actually benefits from caching, because that is where the bill is won or lost.
The second is the default change itself. Teams that pinned Fable 5.1 are moved automatically, so the model writing your code can change without anyone deciding it should. Buyers should expect suppliers to pin and record the model version used on each project, and to re-run tests when it changes.
So what
Model upgrades now arrive weekly, and the commercial question is no longer which model is best but who controls when you move. Treat the model version as part of your delivery contract, in the same way you would treat a framework version. If you want help setting that up, see our AI-assisted development service, or start a project.