Intel LLM-Scaler vllm-0.14.0-b8.2 released with official Arc Pro B70 support
Summary
Intel’s LLM-Scaler vllm-0.14.0-b8.2 adds official support for the Arc Pro B70 GPU, enabling Docker-based large-model inference on Battlemage hardware.
View Cached Full Text
Cached at: 04/22/26, 03:13 PM
Intel LLM-Scaler vllm-0.14.0-b8.2 Released With Official Arc Pro B70 Support
Source: https://www.phoronix.com/news/Intel-LLM-Scaler-vllm-0.14-b8.2
As part of Intel’sLLM-Scalerinitiative for AI inferencing on Intel Arc hardware, out today is their vllm-0.14.0-b8.2 update that includes officially supporting the Arc Pro B70 graphics card.
Intel LLM-Scaler provides the Docker-ized approach for deploying large language models on Intel Arc hardware with a particular focus on latest-generation Battlemage graphics hardware, including multi-GPU configurations as part of theProject Battlematrixinitiative that began last year.
With today’s update for their LLM-Scaler stack with vLLM, they have updated the platform image to intel/llm-scaler-platform:26.18.8.2. The only other listed change is now officially supporting the Intel Arc Pro B70 GPU, the BMG-G31 graphics card that recently debuted with 32GB of vRAM and a sub-$1000 price point.
The IntelArc Pro B70continues running well in tests at Phoronix and I will have out some more benchmarks soon.
The new release is tagged onGitHuband also available via Docker Hub. Though since the release announcement notification, they’ve dropped mentioning the highlights and the Arc Pro B70 support, presumably some release process snafu.
Similar Articles
Intel Arc Pro B70 llama.cpp benchmarks posted
Benchmark results for Intel Arc Pro B70 GPU running llama.cpp with SYCL on Qwen models show 63 tokens per second performance.
@TeksEdge: Solved! Qwen3.6-27B-FP8 is now running on Intel Arc Pro B70! LocalMaxxing shows a working 4× Arc Pro B70 32GB run at ~5…
Qwen3.6-27B-FP8 model is now running on Intel Arc Pro B70 GPUs at ~50 tok/s with a vLLM bug fix, marking a significant milestone for Intel GPU local AI inference.
Show-off Saturday: Intel Arc B140 build.
A showcase of a personal local AI inference build featuring Intel Arc B140 GPUs and custom hardware, running llama.cpp with SYCL back-end on Ubuntu.
Intel Arc Pro b70 box & 3D model
Intel reveals the Arc Pro B70 graphics card box and a 3D model of the card, indicating a forthcoming product launch.
Ling-3.0 (BailingMoE3) lands in llama.cpp mainline - Quick benchmarks on Intel Arc B580
Ling-3.0 (BailingMoE3) is now officially supported in llama.cpp, with benchmarks on Intel Arc B580 showing efficient local inference including 128K context in 12GB VRAM.