Issues

• [Engine] 2x RTX 5060 Ti 16 GB: the one-group batch default from 0.2.1 is 17-32% slower with 2-4 requests; --batch-groups auto restores it

  Niko1221/Strata#4129 · Filter by author Superbobo75 opened on Oct 25, 2026
  2

• [Engine] macOS / Metal: open contributions, rebased on v0.2.6

  Niko1221/Strata#4128 · Filter by author mnott opened on Oct 25, 2026
  2

• [Other] Escalation denied in Plan mode is reported as a rejection by the user

  Niko1221/Strata#4124 · Filter by author WolleGeee opened on Oct 25, 2026

• [Engine] DeepSeek: pull the CPU-lane experts in place and stream the DSpark windows (Project Maya v1.1.28, their PR #476)

  Niko1221/Strata#4123 · Filter by author whitespades opened on Oct 25, 2026

• [Benchmark] Community benchmark of deepseek-v4.1-flash (GSQ-RCO IQ2_XS) on 8x NVIDIA GeForce RTX 3090

  Niko1221/Strata#4122 · Filter by author vikcious opened on Oct 25, 2026

• [Other] Configurable permission presets: a workspace-write mode between Manual and Bypass

  Niko1221/Strata#4120 · Filter by author WolleGeee opened on Oct 25, 2026
  2

• [Engine] Opt-in two-pass QSA prompt selection: 3xTF32 only near the top-k cut-off, the same cells

  Niko1221/Strata#4118 · Filter by author bitaccesscomau opened on Oct 25, 2026

• [Engine] IQ3_S fails to start on a 64 GB-visible Strix Halo (Windows): the expert arena exhausts available RAM (error 1453)

  Niko1221/Strata#4116 · Filter by author qwertie opened on Oct 25, 2026

• [GUI] DeepSeek-V4.1: the conversation context chosen in the app never reaches setup (always --context 8192)

  Niko1221/Strata#4115 · Filter by author beastrobin opened on Oct 25, 2026
  2

• [Engine] Fused prompt experts on a 12.0 card with 70 SMs (RTX 5070 Ti, IQ3_XXS pack, Windows): prompts +11% / +5.6% / +6.7% at 4K / 20K / 98K, 18/18 pairs

  Niko1221/Strata#4110 · Filter by author johnmark13 opened on Oct 25, 2026

• [Benchmark] RTX 3090 Ti, 24 GiB VRAM, 64GB RAM qwen3.8-flash-next-iq3_s

  Niko1221/Strata#4109 · Filter by author AlBundy33 opened on Oct 25, 2026

• [Engine][TUI] engine 0.1.50 did not start

  Niko1221/Strata#4108 · Filter by author almirjakubaitis opened on Oct 25, 2026
  1

• [Benchmark] Strix Halo 128 GB: v0.2.4 stable and faster than llama.cpp, but greedy output diverges more and image text misread

  needs-format
  Niko1221/Strata#4106 · Filter by author thregit opened on Oct 25, 2026
  1

• [Engine] --ple-io ram faults the 26.8 GiB PLE table in with a per-page userspace toucher: 310 s (~90 MB/s) where one MADV_POPULATE_READ runs at the drive's 979 MB/s

  Niko1221/Strata#4104 · Filter by author chenyb999-zhcn opened on Oct 25, 2026

• [Benchmark] NVIDIA GeForce RTX 5060 Ti 16GiB, Intel Xeon E5-2696 v3, 64GB DDR3, PCIe gen3 x8, qwen3.8-flash-next-iq2_xs

  Niko1221/Strata#4103 · Filter by author chuSS opened on Oct 25, 2026

• [Other] how to convert "local" settings (strata-*.json) to launch-presets (strata launch)?

  Niko1221/Strata#4102 · Filter by author AlBundy33 opened on Oct 25, 2026
  2

• NPU

  needs-format
  Niko1221/Strata#4101 · Filter by author Gobblerpl opened on Oct 25, 2026
  1

• [Engine] openapi models endpoint not fully implemented

  Niko1221/Strata#4100 · Filter by author pessimism opened on Oct 25, 2026
  2

• [Engine] --prefill-head-alias: every request fails with "native head upload: out of memory"

  Niko1221/Strata#4098 · Filter by author PierpaoloPernici opened on Oct 25, 2026
  2

• [Other] Can't update from command line

  Niko1221/Strata#4097 · Filter by author xlazom00 opened on Oct 25, 2026

• [Benchmark] RTX 3090 Ti, 24 GiB VRAM, 64GB RAM swift-1.5-iq2_xs

  Niko1221/Strata#4095 · Filter by author AlBundy33 opened on Oct 25, 2026

• [Benchmark] RTX 3090 Ti, 24 GiB VRAM, 64GB RAM qwen3.8-flash-next-iq2_xs

  Niko1221/Strata#4094 · Filter by author AlBundy33 opened on Oct 25, 2026

• [Benchmark] RTX 3090 Ti, 24 GiB VRAM, 64GB RAM qwen3.8-flash-next-iq3_xxs

  Niko1221/Strata#4093 · Filter by author AlBundy33 opened on Oct 25, 2026

• [Benchmark] RTX 3090 Ti, 24 GiB VRAM, 64GB RAM, swift-1.5-iq3_xxs

  Niko1221/Strata#4092 · Filter by author AlBundy33 opened on Oct 25, 2026

• [GUI] [TUI] setup fails on two NVIDIA gpus system (RTX3080ti + RTX5070) all versions 0.2.x.y

  Niko1221/Strata#4091 · Filter by author AndrewDBN opened on Oct 25, 2026