Intel has released LLM-Scaler vllm-0.14.0-b8.2, updating their Dockerized LLM inference stack for Intel Arc hardware. The key addition is official support for the Arc Pro B70 GPU (BMG-G31), which features 32GB of VRAM at a sub-$1000 price point. The platform image has also been updated to intel/llm-scaler-platform:26.18.8.2. The release is available on GitHub and Docker Hub as part of Intel's broader Project Battlematrix initiative for multi-GPU LLM deployments on Battlemage-generation hardware.
Sort: