evals(community): OsaurusAI/Qwen3.8-Flash-Next-JANG_4M on Apple M5 Max

[jax-0n-git]

Crowdsourced model-compatibility run (see COMMUNITY_EVALS.md).

One contribution file; no other changes.

Environment: chip=Apple M5 Max · RAM 128GB · osVersion=26.6.2 · commit=bebc6f3 · judge=self-judge · catalogHash=f2f4dab21df21cfe · contributor=jax-0n-git

[jjang-ai]

Cross-family audit: Flash-Next is behaviorally promising here (AgentLoop 27/33, loop-free 96.97%, 41.36 tok/s), but its 89,945.58 MB peak physical footprint is a hard memory concern and the recorded commit predates the legacy-Hermes/Qwen parser correction now in #2543. This aggregate also drops QSA/MTP token work, exit origins, and cache totals. PR #2546 preserves those fields. Required current-head rerun: parser-corrected bundle path; MTP off/auto/d1/d2/d3 when supported; QSA/MTP verification/accepted/rejected/AR; tool-result continuation; 10-turn, Stop/restart/disk restore; exact TTFT/tok-s/footprint; visible Release-app proof.