moonshine Releases
49 releases of moonshine-ai/moonshine
- v0.1.5
Aug 24, 2026
- v0.1.3
Aug 18, 2026
- v0.1.2
Aug 13, 2026
- b10327
Fixes a critical bug in CUDA kernel launches for quantized copy operations where incorrect thread/block sizing could lead to incorrect results.
Aug 8, 2026
- b10326
Improves the accuracy of timing measurements in the Text-to-Speech (TTS) pipeline by properly accounting for the vocoder pass execution time.
Aug 7, 2026
- b10322
Optimizes SYCL backend performance for SSM convolution operations by coalescing window loads, resulting in significant speedups on Arc Pro B70.
Aug 7, 2026
- b10321
Fixes a bug in the Metal backend where Normalization (NORM) and RMS_NORM would produce incorrect results for specific row lengths due to partial SIMD group handling.
Aug 7, 2026
- b10319
Fixes an issue in Multi-Task Multi-Modal (MTMD) processing where the longest_edge setting was incorrectly ignored when minimum or maximum pixel constraints were applied.
Aug 7, 2026
- v0.1.1
Aug 7, 2026
- v0.1.7Version 0.1.7
This release focuses on performance optimizations for PPTX chart conversion and fixes for LaTeX rendering, PPTX image handling, and documentation.
Jul 29, 2026
- v0.1.0
Jul 27, 2026
- v1.9.0
This release focuses on stabilizing internal state management and TypeScript infrastructure. It introduces a shared state machine micro-kernel for chat streams, removes deprecated skills, and upgrades to TypeScript 6 while maintaining backward compatibility.
Jul 27, 2026
- v0.0.73
Jul 24, 2026
- v0.0.71
Jul 23, 2026
- v1.8.0
A release focused on stability, performance, and security improvements. It includes UI redesigns for the plugins page, enhanced crash reporting capabilities (including OOM handling), and the addition of Korean language support.
Jul 20, 2026
- v0.0.69
Jul 16, 2026
- v0.0.68
Jul 11, 2026
- v0.0.67
Jul 10, 2026
- v0.0.66
Jul 10, 2026
- v0.9.0
A major version update introducing support for LFM 2.5 models (dense and multimodal), multi-model planning capabilities, and significant performance optimizations across CPU and Metal backends.
Jul 7, 2026