Mixed-precision quantization for Apple Silicon, text-only mode (vision tower stripped). Recommended text-only version — best quality-to-size ratio. Quantized from lordx64/Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled using oMLX’s oQ6 algorithm (sensitivity-aware mixed-precision quantization). Drop-in compatible with OpenAI clients including Claude Code (with custom backend), AstrBot, Open WebUI, LibreChat, and Continue.dev. Stripping the vision tower offers practical advantages for text-only workflows: — ~10% faster generation vs the VLM equivalent (no vision compute path overhead) — ~2 GB less peak memory — Simpler deployment — no need for vision processor configs or image preprocessing dependencies If you only feed text into your model, this version is strictly better than the VLM variant. The jump from oQ4 to oQ6 yields a meaningful improvement in retained world knowledge and entity-level recall, while remaining 8 GB smaller than oQ8 with marginal observable quality difference. For most users this is the right balance. Inherits the Claude reasoning distillation: the model uses … tags to structure its chain-of-thought before producing the final response. Best for: -…
Модальности:
Генерация текста
Области применения:
Логика и рассуждение Диалог / чат
Задача: Генерация текста
Автор: wangkezun
Теги: mlx, qwen3_5_moe, apple-silicon, oq, oq6, quantized, 6-bit, moe
Лайков: 3 | Загрузок: 187
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
