Release · Oct 6, 2026 · Glyd
Glyd v0.29.1: more newer models, less GPU memory
Glyd v0.29.1 improves startup and tool calls for newer model families, compresses more loaded weights, and fixes correctness for mixture-of-experts models.
Glyd v0.29.1 is out. It improves newer-model startup, compresses more loaded weights while preserving their original values, and fixes correctness for several mixture-of-experts models.
Install or update with the same commands:
curl -LsSf https://getglyd.com/install.sh | sh
glyd run Qwen/Qwen3.5-9BGemma 4 and Nemotron 3 Nano start with glyd run. Context sizing, tool calls, reasoning output, and suggestions for a smaller model are corrected for newer model families.
More of the weights that v0.29.0 left uncompressed are now compressed in GPU memory. Every compressed weight reads back bit for bit. Saved models can also hold additional weights compressed for a smaller download; support for serving those saves remains limited.
The license and installation requirements are unchanged. Full details are in the release notes and the changelog. Earlier measurements and their sources remain on the benchmarks page.