Describe the bug
Environment:
Hardware: Mac Studio / Mac Pro (M2 Ultra, 196GB Unified Memory)
Model: Qwen 3.5 122B (Versions: 122ba10b Q4 and Q8)
Clients: OpenCode or Cline

Settings: Official Qwen recommended parameters

Description
A consistent crash occurs when the input context length exceeds approximately 42k tokens.

Behavior under 42k: The model performs normally with fast inference and stable memory usage.

Behavior above 42k: The model runner crashes immediately upon reaching the 42k threshold.

Consistency: This issue persists across different quantization levels (both Q4 and Q8) and different integration tools (OpenCode or Cline), suggesting a potential issue with the KV cache management or the context window implementation for this specific model version on Apple Silicon.

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: [e.g. iOS]
• Browser [e.g. chrome, safari]
• Version [e.g. 22]

Smartphone (please complete the following information):

• Device: [e.g. iPhone6]
• OS: [e.g. iOS8.1]
• Browser [e.g. stock browser, safari]
• Version [e.g. 22]

Additional context
Add any other context about the problem here.