Issues

• Experiment: reduce GPU queue overhead with coarser tasks and static fusion

  incoai/splash#760 · Filter by author linson007 opened on Nov 12, 2026

• Experiment: improve draft preparation while preserving attention parallelism

  incoai/splash#759 · Filter by author linson007 opened on Nov 12, 2026

• Concurrent long-context sessions take turns: a Start's lane-state claim waits on the host margin

  incoai/splash#758 · Filter by author avifenesh opened on Nov 12, 2026

• Reduce CommandGraph parameter and literal-binding allocations

  incoai/splash#755 · Filter by author linson007 opened on Nov 12, 2026

• Measure and remove the draft attention layout-copy dispatch

  incoai/splash#754 · Filter by author linson007 opened on Nov 12, 2026

• serve: listen on a UNIX socket (--host unix:PATH)

  incoai/splash#743 · Filter by author JayToltTech opened on Nov 11, 2026

• Feature request: GLM-5.3-Flash family (glm5-next) with SSD expert streaming

  incoai/splash#740 · Filter by author aidendrake9999 opened on Nov 11, 2026
  2

• M5 Max (T6050) total-system watchdog panic under ~112 GB engine pool: 'no checkins from watchdogd in 91 s' (macOS 27.2/26B123)

  incoai/splash#735 · Filter by author brainnoms opened on Nov 10, 2026
  1

• Keep a reusable state at the end of a generated reply, so follow-up turns don't re-prefill it (hybrid GDN models)

  incoai/splash#732 · Filter by author JSummerer opened on Nov 9, 2026

• json_schema responses: the grammar allows 64 whitespace characters before the JSON value, and a client cannot ask for compact output

  incoai/splash#730 · Filter by author tarrific opened on Nov 9, 2026

• Experiment: qualify shared-prefix tree verification and conservative admission

  incoai/splash#728 · Filter by author linson007 opened on Nov 9, 2026

• Request DFlash2-FP8 support for m6

  incoai/splash#725 · Filter by author nailarch opened on Nov 8, 2026

• New Dspine Drafter By Tencent

  incoai/splash#724 · Filter by author kg55555 opened on Nov 8, 2026

• 1.6.0: persistent cache writes ~2 TB/day to the internal SSD; about half of state write-behinds unneeded

  incoai/splash#720 · Filter by author dmkjr opened on Nov 8, 2026
  1

• Contributor questions: DFlash2 draft training data, checkpoint contributions, and which open work to review

  incoai/splash#714 · Filter by author linson007 opened on Nov 7, 2026

• Design: cost-aware allocation across compact32 and wide64 verification

  incoai/splash#709 · Filter by author linson007 opened on Nov 7, 2026
  2

• Research: compact mixer verification projections after FFN packing

  incoai/splash#708 · Filter by author linson007 opened on Nov 7, 2026
  2

• Research: selected-prefix FFN packing for bounded eight-request decode

  incoai/splash#707 · Filter by author linson007 opened on Nov 7, 2026
  2

• 1.6.0 has major cache-loss issues not seen in 1.5.0 on M5 Max

  incoai/splash#706 · Filter by author tomchk opened on Nov 7, 2026
  10

• Sundry macOS App and Web UI Improvements

  incoai/splash#703 · Filter by author jiiim opened on Nov 6, 2026

• Feature: Session-level KV cache release APIs

  incoai/splash#701 · Filter by author cuijs2026 opened on Nov 6, 2026

• Serve record stays "ready" through an in-process engine relaunch, so a supervisor cannot tell the relaunch from a runaway load

  incoai/splash#700 · Filter by author pcvelz opened on Nov 6, 2026

• [Feature/Robustness] Complete agentic workflow support in Responses API: tolerant input items & compaction output handling

  incoai/splash#695 · Filter by author hometrix opened on Nov 6, 2026

• Splitting the mixers' projections with the Neural Engine

  incoai/splash#690 · Filter by author fishy-ops opened on Nov 6, 2026

• [feature] Support Ripple-65B-Preview (Dense Model) as a Splash package

  incoai/splash#689 · Filter by author kawaiivoldemort opened on Nov 5, 2026
  1