[Bug] Claude Code fails to generate grammatically correct Swedish B1-level sentences consistently

Bug Description
cant put together a b1 level sentnec after 5 tries . A reasonable performance expectation for an AI tool marketed at Claude’s level when generating Swedish:
• Grammatical accuracy: 95%+ on standard sentences
• Word order: correct in nearly all standard structures
• Vocabulary: appropriate for context
• Article and gender agreement: 90%+
• Verb conjugation: 95%+ Basic correct sentence generation is well within the baseline a competent
AI tool should deliver. Native-level nuance (cultural register, regional variation, complex idiom) is where actual edge-case errors would be expected.Systematic failure at correct sentence-level construction in Swedish after three and a half hours of corrections indicates the tool is not delivering at the foundation level, not at the edge of capability.
Having to ask the tool to redo a B1-level sentence (basic intermediate, well within standard capability range) five times before getting an acceptable output is documented evidence of this foundational failure.
This is either severe degradation for the consumer tier, a specific bug in non-English language handling, or token throttling affecting output quality. Any of these constitutes failure to deliver the advertised product

Environment Info

• Platform: darwin
• Terminal: Apple_Terminal
• Version: 2.1.195
• Feedback ID: efb72b3a-301a-40fc-88cd-83140e462d66

Errors

[]