Skip to content

vMLX 1.6.16

Choose a tag to compare

View all tags
@jjang-ai jjang-ai released this 23 Jul 03:53
v1.6.16
3ee1f64

vMLX 1.6.16 is an emergency parser, cache, and bundled-runtime checkpoint.

Highlights:

  • Ships the KaTeX-backed Electron renderer path for chat and reasoning surfaces.
  • Retains scoped reasoning_content, visible-output, and tool-call streaming fixes/proofs for the 1.6.16 parser/cache gate.
  • Hardens packaged engine provenance so signed app sessions use bundled Python/vmlx_engine instead of stale development or system virtualenvs.
  • Retains scoped block-disk/L2/TurboQuant cache proofs, including Paged-Off SSD/TQ reuse paths covered by current reruns.
  • Bundled app artifacts use vmlx_engine 1.6.16 and local jang_tools 2.5.34 with bundled-source parity verification.

Signed and notarized app artifacts are distributed from the companion MLXStudio release:
https://github.com/jjang-ai/mlxstudio/releases/tag/v1.6.16

Known follow-up:
This is a public emergency checkpoint, not full closure of every retained family/media/gateway stress row. Broader family repetition, media/audio/video matrix, exhaustive gateway soak, PyPI JANG publishing, and remaining documented PARTIAL rows remain follow-up work.

Assets 2
Loading