Issues

• [Feature]: MTP drafts in pipelined batch groups (2-GPU layer split, parallel 2)

  Niko1221/Strata#2235 · Filter by author davidbhoward opened on Oct 13, 2026

• Throughput with 4 slots on 2x RX 7900 XTX + an old Xeon: where to go from here?

  Niko1221/Strata#2232 · Filter by author mantovaniluca91 opened on Oct 13, 2026

• Conversation cache falls back to 32,768-spaced checkpoints for a fast client: 13 re-reads of 6.8K-47K tokens (4-17 s TTFT) in 3.5 h while its prompts are pure appends

  Niko1221/Strata#2228 · Filter by author jy02375437-max opened on Oct 13, 2026
  1

• --mtp-hnorm stream beats the default on sm_89 (IQ3_XXS): draft acceptance 82%->86%, decode +3.8%/+2.3% at 6K/70K, neutral at 120K, answers byte-identical

  Niko1221/Strata#2227 · Filter by author jy02375437-max opened on Oct 13, 2026

• Configuration report: RTX 5090, IQ3_S + INT8 KV + BF16 vision, four 192K slots

  Niko1221/Strata#2225 · Filter by author TurboTheTurtle opened on Oct 13, 2026

• Community benchmark: 2x RTX 4070, E5-1630, 32GB, IQ1_M, 64K

  Niko1221/Strata#2219 · Filter by author KanzoHaru opened on Oct 13, 2026

• RTX 5090 + Core Ultra 9 285K，Strata v0.1.43.1 / IQ2_XS，adaptive overlap 开关对照

  Niko1221/Strata#2218 · Filter by author liuhao16 opened on Oct 13, 2026

• Community benchmark on rtx-3050-plus-quadro-p6000-qwen3-8-flash-next-q2-0

  Niko1221/Strata#2214 · Filter by author almirjakubaitis opened on Oct 13, 2026

• Community benchmark on quadro-p6000-qwen3-8-flash-next-iq2-xs

  Niko1221/Strata#2213 · Filter by author almirjakubaitis opened on Oct 13, 2026

• sudden crash on RTX 3070 for long task

  Niko1221/Strata#2212 · Filter by author HiroshiMatsumoto opened on Oct 13, 2026

• Community benchmark on quadro-p6000-qwen3-8-flash-next-q2-0

  Niko1221/Strata#2211 · Filter by author almirjakubaitis opened on Oct 13, 2026

• [Bug]: Windows HIP gfx1031: prompt reading ~30 % slower on the 0.1.43 engine than 0.1.40 (same model and server)

  Niko1221/Strata#2210 · Filter by author huy-sec opened on Oct 13, 2026

• Community benchmark on NVIDIA GeForce RTX 5060 Ti - Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S

  Niko1221/Strata#2202 · Filter by author williamckl opened on Oct 13, 2026

• The issue of execution being interrupted when integrating with Trae and other development tools.

  Niko1221/Strata#2200 · Filter by author lonestonewy opened on Oct 13, 2026
  2

• Turing (sm_75, RTX 20) support: tracking issue for the open PRs

  Niko1221/Strata#2197 · Filter by author rafatxf opened on Oct 13, 2026
  1

• SYCL, Arc A770 (xe): decode intermittently never starts - host waits in Verifier::run on a queue that does not drain, RSS grows until OOM

  Niko1221/Strata#2194 · Filter by author aslater3 opened on Oct 12, 2026

• Crash sometimes serving IQ3_S

  Niko1221/Strata#2192 · Filter by author GarryLai opened on Oct 12, 2026

• 0.1.43.1, Windows: STRATA_ADAPT_OVERLAP=1 stops the first request with "the adaptive tier's copies for layer 45 never landed" (RTX 4070 SUPER); STRATA_ADAPT_SIGNAL=host works

  Niko1221/Strata#2188 · Filter by author LeGeRyChEeSe opened on Oct 12, 2026

• Community benchmark: RTX Pro 4000 Blackwell, qwen3.8-flash-next IQ3_S (bench --long)

  Niko1221/Strata#2182 · Filter by author DrLove73 opened on Oct 12, 2026

• Community benchmark: RTX Pro 4000 Blackwell, qwen3.8-flash-next IQ3_S --prefill auto:32768 --conversation-cache-mib 8192 (bench --long)

  Niko1221/Strata#2181 · Filter by author DrLove73 opened on Oct 12, 2026

• Community benchmark: RTX Pro 4000 Blackwell, qwen3.8-flash-next IQ3_S

  Niko1221/Strata#2180 · Filter by author DrLove73 opened on Oct 12, 2026

• Field report: --resident-experts and layer_split auto` costs nothing on 128 GB ram

  Niko1221/Strata#2177 · Filter by author tonydiep opened on Oct 12, 2026

• Community benchmark: RTX Pro 4000 Blackwell, Swift 1.5 IQ3_XXS --prefill auto:32768 --conversation-cache-mib 8192 (bench --long)

  Niko1221/Strata#2171 · Filter by author DrLove73 opened on Oct 12, 2026

• Community benchmark: RTX Pro 4000 Blackwell, Swift 1.5 IQ3_XXS 1

  Niko1221/Strata#2169 · Filter by author DrLove73 opened on Oct 12, 2026

• v0.1.43.1 and v0.1.43 are slower than v0.1.42, and outputs that are not truncated in 0.1.42 are truncated before completion in v0.1.43 and v0.1.43.1, with no errors in the logs

  Niko1221/Strata#2167 · Filter by author zhang2939 opened on Oct 12, 2026