Issues

• [Feature Request] Native Jev Mode (/v1/systemone Endpoint) for Fast Typed Decisions & Logit Probabilities

  AtomicBot-ai/atomic-llama-cpp-turboquant#84 · Filter by author Psanyi89 opened on Sep 24, 2026

• Eval bug: ROCm backend crashes on first inference (gfx1201) — missing Tensile library & precompiled kernels

  AtomicBot-ai/atomic-llama-cpp-turboquant#83 · Filter by author RenZekta opened on Sep 23, 2026

• Feature Request: please ad support to model architecture: 'qwen4exp'

  enhancement
  AtomicBot-ai/atomic-llama-cpp-turboquant#82 · Filter by author dan64 opened on Sep 16, 2026

• Eval bug: Crashes with "ROCm error: device kernel image is invalid"

  AtomicBot-ai/atomic-llama-cpp-turboquant#81 · Filter by author Black6spdZ opened on Sep 15, 2026

• [Windows] Near-limit VRAM reservation causes silent WDDM shared-memory spill at request time - suggest startup VRAM watermark warning

  AtomicBot-ai/atomic-llama-cpp-turboquant#77 · Filter by author zack0418 opened on Sep 9, 2026

• Eval bug: error loading model: unknown model architecture: 'bailingmoe3' in latest b10269-1.5.1

  bugreleases
  AtomicBot-ai/atomic-llama-cpp-turboquant#69 · Filter by author dineshr93 opened on Aug 9, 2026
  2

  

• Linux Vulkan b10018-1.3.0: Gemma 4 31B output is corrupted; upstream llama.cpp works

  bugVulkan
  AtomicBot-ai/atomic-llama-cpp-turboquant#61 · Filter by author hsjlyj opened on Aug 4, 2026
  1