LocalForge
返回指南
模型 / COMMUNITY REAP BUILD

Qwen3.8 Flash Next REAP-288

模型卡核對於 2026-09-16 · 記憶體適配為估算值；除非有符合你配置的獨立實測，否則速度未知。

也會被搜尋為：qwen3.8-flash-next-reap-288-mlx-4bit

Community-pruned conversion, not the stock checkpoint. The 39 GB model / ~40.57 GB total resident path requires oMLX 0.6.4+ and streams only the sparse PLE table from NVMe. A published M4 Pro 48 GB test measured 26–29 tok/s and recommended 7,680 tokens; do not generalize it to arbitrary models or Macs.

一手來源
總參數176B
估算記憶體~40.57 GBREAP-288 · MLX 4-bit · PLE on NVMe
上下文7.6K recommended on measured 48 GB MacText
授權條款Qwen CommunityMultilingual; post-pruning quality not independently established
01 / 適合
  • 48 GB Mac experiments
  • coding agents
  • model-specific NVMe streaming
02 / 估算適配
Mac Studio M5 Ultra · 96 GB餘量充足MacBook Pro M5 Max · 64 GB勉強適配Mac Studio M3 Ultra · 96 GB餘量充足Mac Studio M3 Ultra · 256 GB餘量充足Apple silicon · 64 GB勉強適配
03 / 部署方式

以下是執行方式範例，不是已驗證的安裝流程。容量估算對應量化檔案，官方倉庫可能是全精度權重；MLX、GGUF 與 TensorRT 需要相符的轉換檔或引擎。

oMLX進階omlx serve --model-dir ~/models --memory-guard-gb 48
注意

Community-pruned conversion, not the stock checkpoint. The 39 GB model / ~40.57 GB total resident path requires oMLX 0.6.4+ and streams only the sparse PLE table from NVMe. A published M4 Pro 48 GB test measured 26–29 tok/s and recommended 7,680 tokens; do not generalize it to arbitrary models or Macs.