[BUG] Explicit model pin claude-fable-5[1m]

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Explicit model pin claude-fable-5[1m] (in both ~/.claude/settings.json and project .claude/settings.local.json) with switchModelsOnFlag: false, yet sessions keep getting server-side-routed to Opus 5 mid-work. No client setting stops it. The trigger appears to be defensive security-remediation content (closing an unauthenticated-access hole in a payments backend — exploit-shaped text, but the intent is to fix it, not exploit it). A deliberate pin shouldn't be silently overridden. Asks: (1) don't override an explicit pin without surfacing that it happened and why; (2) give a real opt-out for content-based re-routing; (3) don't treat defensive security work as offensive.

What Should Happen?

Explicit model pin claude-fable-5[1m] (in both ~/.claude/settings.json and project .claude/settings.local.json) with switchModelsOnFlag: false, yet sessions keep getting server-side-routed to Opus 5 mid-work. No client setting stops it. The trigger appears to be defensive security-remediation content (closing an unauthenticated-access hole in a payments backend — exploit-shaped text, but the intent is to fix it, not exploit it). A deliberate pin shouldn't be silently overridden. Asks: (1) don't override an explicit pin without surfacing that it happened and why; (2) give a real opt-out for content-based re-routing; (3) don't treat defensive security work as offensive.

Error Messages/Logs

Explicit model pin claude-fable-5[1m] (in both ~/.claude/settings.json and project .claude/settings.local.json) with switchModelsOnFlag: false, yet sessions keep getting server-side-routed to Opus 5 mid-work. No client setting stops it. The trigger appears to be defensive security-remediation content (closing an unauthenticated-access hole in a payments backend — exploit-shaped text, but the intent is to fix it, not exploit it). A deliberate pin shouldn't be silently overridden. Asks: (1) don't override an explicit pin without surfacing that it happened and why; (2) give a real opt-out for content-based re-routing; (3) don't treat defensive security work as offensive.

Steps to Reproduce

Explicit model pin claude-fable-5[1m] (in both ~/.claude/settings.json and project .claude/settings.local.json) with switchModelsOnFlag: false, yet sessions keep getting server-side-routed to Opus 5 mid-work. No client setting stops it. The trigger appears to be defensive security-remediation content (closing an unauthenticated-access hole in a payments backend — exploit-shaped text, but the intent is to fix it, not exploit it). A deliberate pin shouldn't be silently overridden. Asks: (1) don't override an explicit pin without surfacing that it happened and why; (2) give a real opt-out for content-based re-routing; (3) don't treat defensive security work as offensive.

Claude Model

Other

Is this a regression?

Yes, this worked in a previous version

Last Working Version

5

Claude Code Version

2.1.216 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Terminal.app (macOS)

Additional Information

Explicit model pin claude-fable-5[1m] (in both ~/.claude/settings.json and project .claude/settings.local.json) with switchModelsOnFlag: false, yet sessions keep getting server-side-routed to Opus 5 mid-work. No client setting stops it. The trigger appears to be defensive security-remediation content (closing an unauthenticated-access hole in a payments backend — exploit-shaped text, but the intent is to fix it, not exploit it). A deliberate pin shouldn't be silently overridden. Asks: (1) don't override an explicit pin without surfacing that it happened and why; (2) give a real opt-out for content-based re-routing; (3) don't treat defensive security work as offensive.