[BUG] Severe quality degradation despite /model showing Opus - possible silent model mismatch #13943

Preflight Checklist

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

What's Wrong?

Since approximately 3 days ago, Claude Code shows severe quality degradation in following basic instructions, despite /model command confirming Opus is active.

Symptoms:
Ignores explicit file targets (e.g., "add to meta.json" → edits ai-research.json instead)
Claims context is lost even though the text was just provided
Repeatedly misquotes user input when asked to repeat it
Doubles down on errors instead of re-reading instructions
Behavior matches reports of silent Sonnet fallback, but /model still shows Opus

Relation to #3434
This may be related to #3434 (silent fallback to Sonnet), but with an additional issue: the /model command does not reflect the actual model being used. The UI/CLI says Opus, but behavior suggests otherwise.

Before (3+ days ago):
Instructions followed correctly
Same workflow worked without issues

After:
Simple tasks fail repeatedly
Must start new chats constantly
Even first message in fresh chat shows degraded behavior

What Should Happen?

If /model shows Opus, responses should reflect Opus-level instruction following.
Request
Please investigate whether there's a mismatch between displayed model and actual model being served.

Error Messages/Logs

Steps to Reproduce

Give clear instruction with explicit file target: "add X to meta.json"
Claude edits wrong file or misinterprets completely
Check /model → shows Opus
Behavior does not match Opus capabilities

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

2.0.69

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

VS Code integrated terminal

Additional Information

No response