Strata v0.1.40.2 is an open-source C++ inference engine that allows users to run the Qwen3.8-Flash-Next model on consumer gaming PCs without cloud data transmission, now with added support for Intel Arc GPUs on Linux and improved performance across NVIDIA and AMD hardware. This update enhances speed, particularly in multi-GPU setups and low-RAM environments, while ensuring that output remains byte-identical to the previous version. Notable improvements include faster processing of long documents and a streamlined installation process, with a browser app for easy interaction. The project, maintained by a developer named Niko1221, benefits from community contributions and is part of a broader trend towards enabling complex AI models to run on personal hardware

Strata v0.1.40.2 Adds Intel Arc Support and Speed Gains for Local LLMs

Strata is a free, open-source C++ inference engine that runs the ~125-billion-parameter Qwen3.8-Flash-Next model on consumer gaming PCs without sending data to the cloud. Its new v0.1.40.2 release, out today, officially adds Intel Arc GPU support on Linux while lifting speed across NVIDIA and AMD hardware. The biggest gains show up in multi-GPU splits, machines with less RAM than the model, and long-document workflows, and default output stays byte-identical to 0.1.40. Kept on GitHub by Niko1221, the MIT-licensed project keeps shipping multiple releases per week, leaning on community contributions to test hardware it can't own itself.

Strata v0.1.40.2 Adds Intel Arc Support and Speed Gains for Local LLMs @ Linux Compatible