Strata 0.2.15 Bug-Fix Update: Removes llama-server Dependency, Fixes Drafting and Downloads

Published by

Strata 0.2.15 is a bug-fix update for the inference engine that operates 177B-parameter models on inexpensive GPUs, ensuring that user models, chats, and settings remain unchanged. This release eliminates the external llama-server dependency, allowing models to load directly on Strata's engine and improving the system's stability. Key fixes include the ability to skip unnecessary downloads during drafting, a resolution for the Qwen3.6-35B-A3B first-run install issue, and enhanced terminal functionality that recognizes AMD cards on hybrid systems. The update is now available on GitHub, providing a smoother experience without requiring any additional account or cloud services



Strata 0.2.15 Bug-Fix Update: Removes llama-server Dependency, Fixes Drafting and Downloads

Strata 0.2.15 is a bug-fix update for the inference engine that runs 177B-parameter models on cheap GPUs, and it keeps your models, chats, and settings intact. The release drops the external llama-server dependency so models load on Strata's own engine instead, removing a moving part that quietly drifted out of sync. It fixes drafting by skipping the roughly 5 GB MTP download when you turn it off, saving that choice across restarts, and repairing the Qwen3.6-35B-A3B first-run download that previously died after four seconds. The terminal tool now picks up AMD cards on hybrid NVIDIA plus AMD machines, remembers your drafter choice over multiple restarts, and names missing Linux libraries with the exact command to install them.

Strata 0.2.15 Bug-Fix Update: Removes llama-server Dependency, Fixes Drafting and Downloads @ Linux Compatible