Issues

• Prefill panics above ~128k tokens: routing ids become garbage (qwen4exp, Vulkan)

  RustingStudio/RustingStrata#25 · Filter by author ksergk opened on Nov 10, 2026
  3

• [Idea]: Allow calibrate to accept multiple devices like serve does

  RustingStudio/RustingStrata#23 · Filter by author Darksecond opened on Nov 10, 2026
  2

• [Bug]: 9070 XT on Windows maxComputeSharedMemorySize incorrectly using 32768 causing failure to load model

  RustingStudio/RustingStrata#20 · Filter by author BrightCandle opened on Nov 10, 2026
  1

• [Idea]: possible support for gemma4 26b a4b?

  RustingStudio/RustingStrata#19 · Filter by author beyondzvincent opened on Nov 10, 2026
  3

• [Idea]: MCP support and options for the GUI interface

  RustingStudio/RustingStrata#18 · Filter by author KotPrezesOfficial opened on Nov 9, 2026
  1

• [Idea]: Deepseek support?

  Idea
  RustingStudio/RustingStrata#12 · Filter by author luckydevil13 opened on Nov 7, 2026
  13

• RX 6700 XT / RADV: Vulkan works, slow 30K Qwen3.6 prefill

  AMDBenchmark
  RustingStudio/RustingStrata#10 · Filter by author hermeticvm opened on Nov 6, 2026
  3

• RX 9070 XT (Vulkan/RADV), Flash-Next IQ2_XS: decode 0.8 tok/s, thousands of "GPU waits for the CPU expert worker that timed out"

  AMDBenchmarkbugTesting
  RustingStudio/RustingStrata#3 · Filter by author charmoky opened on Nov 5, 2026
  14