spb/zyquo-local Public MIT

1<!--2  VERIFICATION.md3  Zyquo Local45  Author: Simon-Pierre Boucher6  Mail: contact@spboucher.ai7-->89# Phase 7 — Local Model Verification1011Run on 2026-07-30 with `ZyquoLocal --verify` on the build machine12(M5 Max, 48 GB, macOS 27.0 beta). Per model: download through the app's own13`DownloadManager`, directory validation, load, deterministic generation14(temperature 0, "Reply with exactly: OK" → replied "OK"), multi-turn context15carry-over ("vermilion" recall), streaming-cancellation test (stream stops in16≤0.4 s after cancel), unload with verified memory release (MLX active memory17drops to KBs).1819## End-to-end results — 6 models across architectures2021| Model | Download | Load | Generate | Multi-turn | Cancel | Unload | tok/s | TTFT |22|---|---|---|---|---|---|---|---|---|23| Qwen3-0.6B-4bit | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 109.4 | 0.08s |24| Llama-3.2-1B-Instruct-4bit | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 222.8 | 0.74s |25| SmolLM3-3B-4bit | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 97.4 | 0.11s |26| gemma-3-4b-it-qat-4bit | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 58.3 | 1.08s |27| Qwen2.5-Coder-7B-Instruct-4bit | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 112.5 | 0.42s |28| DeepSeek-R1-0528-Qwen3-8B-4bit | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 61.3 | 0.08s |2930Architecture coverage: qwen3, llama, smollm3, gemma3, qwen2 (coder), R131distill (reasoning, `<think>` streaming). Memory release examples:324.65 GB → 20 KB (R1-8B), 2.63 GB → 13 KB (gemma-3-4b).3334Per the protocol, test models except the smallest (Qwen3-0.6B-4bit, kept for35ongoing dev) were deleted afterward to reclaim disk.3637## Featured catalog dry-verification — 30/30 ✅3839Every repo in `docs/MODELS.md` / `ModelCatalog` verified against the live40Hub: repo exists, required files listed, total size within 10 % of the41catalog value (all matched to the hundredth of a GB; no gated repos, no42missing files, no corrections required — `MODELS.md` and `ModelCatalog`43remain in sync).4445**VERDICT: ALL GREEN** (`--verify` exit 0)46