Anthropic · 2026-09-01 · major
Claude Code 2.1.257 — Claude Fable 5.1 becomes the default Fable model
Claude Code 2.1.257 makes Claude Fable 5.1 its default Fable model, with a 1M-token context and $0.25 per million cache reads. Auto mode also stops auto-approving cloud metadata-credential fetches, egress evasion and cross-tenant reach.
Claude Code 2.1.257 switches its default Fable model to Claude Fable 5.1 and stops auto mode waving through container-escape moves.
Quick facts
| Maker | Anthropic |
|---|---|
| Version | 2.1.257 |
| Released | 1 September 2026 |
| Default Fable model | Claude Fable 5.1 (claude-fable-5-1) |
| New auto-mode rule | Containment Escape |
| New settings | timeFormat and timeZone for the turn-end clock and transcript timestamps |
| Install | npm i -g @anthropic-ai/claude-code |
What is it?
Claude Fable 5.1 is now the default Fable model in Claude Code 2.1.257, bringing a 1M-token context and $0.25 per million cache reads to the CLI. The release also adds a Containment Escape rule to auto mode, and timeFormat and timeZone settings that control the turn-end clock and transcript timestamps, including a strftime pattern option.
How does it work?
The Containment Escape rule sits inside auto mode's approval logic, so cloud metadata-credential fetches, egress evasion and cross-tenant reach stop being auto-approved unless the environment marks them expected. A separate one-time prompt fires before the first file read outside the working directories. Two smaller controls land alongside: CLAUDE_CODE_SUBAGENT_MODEL_FORCE pins every subagent to one model, and `s` in /effort scopes an effort change to the current session.
Why does it matter?
Auto mode is the setting people leave running unattended, so what it approves silently is the whole security question. Reading credentials from a cloud metadata endpoint is a standard first step in a container escape, and 2.1.257 stops treating it as ordinary work. On cost, moving the CLI's default Fable model forward cuts cache reads to a quarter of the previous price without changing input or output rates.
Who is it for?
Claude Code users and platform teams
Frequently asked questions
- What does the Containment Escape rule block in Claude Code auto mode?
- The Containment Escape rule in Claude Code 2.1.257 stops auto mode from approving cloud metadata-credential fetches, egress evasion, and cross-tenant reach on its own. Those actions now need approval unless your environment marks them as expected. Fetching credentials from a cloud metadata endpoint is a standard opening move in a container escape, so it no longer counts as routine.
- How do I force every subagent onto one model in Claude Code?
- Claude Code 2.1.257 adds the CLAUDE_CODE_SUBAGENT_MODEL_FORCE environment variable. Setting it applies CLAUDE_CODE_SUBAGENT_MODEL, or the main model when that is unset, to every subagent, ignoring both per-spawn model choices and the model field in an agent definition. Without it, an agent definition's own model still takes precedence.
- Can I change effort for just one Claude Code session?
- Yes. Claude Code 2.1.257 adds an `s` option in /effort that changes the effort level for the current session only, matching how /model already works. That lets you raise or lower effort for a single piece of work without editing the setting your other sessions read.
- Can Claude Code read files outside my working directory?
- Claude Code 2.1.257 adds a one-time prompt in auto mode before the first file read outside the working directories, so an outside read is a decision rather than a default. The prompt offers the option to block such reads, which sets permissions.blockReadsOutsideWorkingDirectories and stops them for good.
Try it
npm i -g @anthropic-ai/claude-code@2.1.257