Claude Code v2.1.257 adds Fable 5.1 model, containment-escape rule, and subagent model forcing
cliClaude Code CLI 2.1.257 ships Claude Fable 5.1 (`claude-fable-5-1`) as the new default Fable model with a 1M context window ($10/$50 per Mtok, $0.25/Mtok cache reads), a new auto-mode Containment Escape rule that blocks cloud metadata credential fetches and cross-tenant reach, `CLAUDE_CODE_SUBAGENT_MODEL_FORCE` to override every subagent's model, time-format/timezone settings, an out-of-workdir file-read consent prompt, and dozens of subagent, MCP, sandbox, and background-ses
An agent can now run on Claude Fable 5.1 with a 1M-token context in Claude Code, force a single model across all subagents via `CLAUDE_CODE_SUBAGENT_MODEL_FORCE`, and rely on a Containment Escape rule that blocks cloud-metadata credential exfiltration in auto mode.
Notable: default Fable model bump to 5.1 with 1M context; new containment-escape guardrail in auto mode; subagent model force env var; broad reliability fixes across background sessions, MCP, and sandbox rules.
npm i -g @anthropic-ai/claude-code@2.1.257npm i -g @anthropic-ai/claude-code@2.1.257
# Use the new default Fable 5.1 model with a forced subagent model
export CLAUDE_CODE_SUBAGENT_MODEL=claude-fable-5-1
export CLAUDE_CODE_SUBAGENT_MODEL_FORCE=1
claude --model claude-fable-5-1 \
--add-dir ./src \
-p "Refactor the auth module and delegate tests to a subagent"
# Block reads outside working directories in auto mode
claude config set permissions.blockReadsOutsideWorkingDirectories true