v1.16.24 @magnitudedev/cli@0.2.5
low risk
nilbuild/developer-roadmapOct 3, 2026 → Oct 3, 20260 days between releases
What's New
Optimizes inference performance for MoE models and accelerates prompt processing on modern Mac hardware using GPU tensor operations.
New Features
- Multi-token prediction speedups for MoE models
- GPU tensor operations for prompt processing on M5+ Macs
- Metal thread limit compatibility fixes
- Vulkan shader compilation error fixes
- Optimized multi-token prediction on NVIDIA GPUs
Improvements
- Speed up generation for mixture-of-experts models
- Speed up prompt processing on M5 and later Macs
- Speed up multi-token prediction on NVIDIA GPUs
Bug Fixes
- Fix for Metal thread limit errors on M1 and M2 Macs
- Fix for Vulkan shader compilation errors on various GPUs
Upgrade Notes
No configuration changes or migration steps are mentioned in the release notes; this appears to be a performance and stability patch.
The Two Releases
v1.16.24
v1.16.24
Oct 3, 2026· 19 days ago
Focuses on inbox inspection tools, session nesting, UI refinements, and a restructuring of the skill installation files.
Full release notes@magnitudedev/cli@0.2.5
@magnitudedev/cli@0.2.5
Oct 3, 2026· 19 days ago
Optimizes inference performance for mixture-of-experts models and accelerates prompt processing on modern Mac hardware using GPU tensor operations.
Full release notes