Issues

• Benchmark: GLM-5.3-Flash on the Metal backend, Mac Studio M3 Ultra 512 GB (Maya-L IQ3_S and unsloth UD-Q4_K_XL)

  mw00/project-maya#986 · Filter by author soulmachine opened on Nov 14, 2026

• v1.1.62: Windows CUDA build of maya-ds4 fails - ds4_rhelper.inc uses usleep/shm_unlink outside the _WIN32 guard

  mw00/project-maya#983 · Filter by author pietroMastro92 opened on Nov 14, 2026
  1

• Speed report: 1× Tesla V100 PCIE 32GB, DeepSeek-V4.1-Flash-1017 Maya-XS v2

  speed-report
  mw00/project-maya#978 · Filter by author wairentiun opened on Nov 13, 2026

• Proposal: Integrate XeMaya's native Intel Arc model paths and optimization R&D

  mw00/project-maya#977 · Filter by author Jumbomuffin777 opened on Nov 13, 2026

• GLM-5.3 (744B): serve KV streaming on glm-dsa - is the L1-6 path meant to work on v1.1.61?

  mw00/project-maya#975 · Filter by author pietroMastro92 opened on Nov 13, 2026
  1

• Add Inkling-2-Small as a new model

  enhancement
  mw00/project-maya#974 · Filter by author kabachuha opened on Nov 12, 2026
  1

• If model dir is on separate disk as /data3 , vision download will default to root / dir

  bug
  mw00/project-maya#972 · Filter by author uris-code1 opened on Nov 12, 2026

• Speed report: DeepSeek EXL3 2.25 bpw on 2x RX 7900 XTX + Radeon 8060S, 1M context: 69.4 tok/s decode, 1,801 tok/s prefill at 120K (v1.1.56 + #867 + #904 + #945 + #967 + #968)

  speed-report
  mw00/project-maya#969 · Filter by author pepuscz opened on Nov 12, 2026
  1

• v1.1.60 HIP build fails on gfx1151 — system /usr/include/hip headers picked over venv ROCm

  amdbuginstall
  mw00/project-maya#965 · Filter by author queensone opened on Nov 12, 2026
  1

• CPU update 1.31x

  speed-report
  mw00/project-maya#961 · Filter by author uris-code1 opened on Nov 11, 2026
  1

• maya 1.1.54 , Deepseek maya-s : 70 tokens/s ! （ai max 395 -128g RAM, rtx 5000 72G VRAM eGPU, CachyOS）

  speed-report
  mw00/project-maya#957 · Filter by author williamShaw-xiao opened on Nov 11, 2026
  1

• DeepSeek --tp 1 on 2x RTX A4000: since v1.1.58 (#697) the PCIe lane stays on and decode is ~5% slower than with MAYA_DS4_STREAM=0

  bug
  mw00/project-maya#956 · Filter by author speja969 opened on Nov 11, 2026
  2

• [DeepSeek-V4.1 EXL3] Engine v1.1.57 hangs right after the prompt rings — never runs a layer (SM80+SM89 mixed, 8 GPUs)

  bugneeds-repro
  mw00/project-maya#955 · Filter by author yunix-intel opened on Nov 11, 2026

• high idle CPU usage, #685 incomplete fix

  bug
  mw00/project-maya#951 · Filter by author vbooka1 opened on Nov 11, 2026

• GPU load estimates may not be accurate on DS4.

  question
  mw00/project-maya#948 · Filter by author bob-u opened on Nov 10, 2026

• [Feature] Control vector support for GLM

  enhancement
  mw00/project-maya#944 · Filter by author kabachuha opened on Nov 10, 2026

• Tesla v100 running Kimi-k3.1-pro with Llama.cpp "2.03 t/s" we need Maya version of this.

  enhancementmodel
  mw00/project-maya#936 · Filter by author gngglobetech opened on Nov 9, 2026
  1

• DeepSeek: a long prompt freezes the running answers for its whole read (4.3 s at 4K) - a pause between layers takes it to 0.4 s, the same bits

  enhancement
  mw00/project-maya#930 · Filter by author linus-aegu opened on Nov 9, 2026
  5

• DeepSeek on 8x RTX 3090 (TP2, EXL3): new-text decode and prefill bound by Engram row reads - 256 MB row cache, ZFS recordsize, DSpark cut (decode 47 -> 69 tok/s, fresh-text prefill 520 -> ~1,500)

  enhancement
  mw00/project-maya#926 · Filter by author olobarius opened on Nov 8, 2026
  3

• Reuse an existing DeepSeek-V4.1 DSpark GGUF without duplicating its weights

  enhancement
  mw00/project-maya#919 · Filter by author mercant1982 opened on Nov 8, 2026
  3

• Project Maya 1.1.56 - report of 2026-11-08 14:45 - crash upon given prompt

  bugneeds-reprowindows
  mw00/project-maya#918 · Filter by author helbitz opened on Nov 8, 2026
  2

• Please optimize for sm_90 GPUs.

  enhancementneeds-repro
  mw00/project-maya#917 · Filter by author 45x929 opened on Nov 8, 2026
  1

• [Feature] support amx cpu

  enhancement
  mw00/project-maya#916 · Filter by author 45x929 opened on Nov 8, 2026
  1

• Illegal memory access when running modified DeepSeek V4.1 Flash 1017

  bugneeds-repro
  mw00/project-maya#907 · Filter by author threefcata opened on Nov 7, 2026
  4

• Possible bug experienced while trying to get Maya-S24 run on dual RX 9070 16GB for the first time

  amdbugneeds-repro
  mw00/project-maya#899 · Filter by author bob-u opened on Nov 6, 2026
  2