Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Python
"""Vendored base model implementations (licensing in NOTICE).
* ``qwen3_next.py`` — mlx-lm 0.31.0's Qwen3.6-35B-A3B model, byte-identical
to the upstream file except the import block, which was rewritten to
absolute ``mlx_lm.models`` paths so the file can live outside the
mlx_lm package. Pinned to mlx-lm==0.31.0 by pyproject.
* ``bailing_hybrid.py`` — the ling 3.0 backbone vendored from the
production deployment, with the prerouter machinery renamed from the
legacy identifiers (``*_gate*`` -> ``prerouter*``) and no other
changes.
"""