ASUS unveils UGen Intel Arc Pro B70 32GB for local AI inference
ASUS announced the UGen Intel Arc Pro B70 32GB, an AI accelerator designed for local agentic AI workloads on desktops and workstations. The company is positioning the card around high-memory local inference as large language models and multi-agent workloads move beyond cloud-only deployment.
The accelerator includes 32GB of ECC-protected GDDR6 memory, which ASUS says is intended to help keep demanding AI workloads on the local system rather than offloading data to host memory or the cloud.
ASUS said the card can scale to 64GB of VRAM in dual-card systems and 128GB in four-card configurations, targeting use cases ranging from independent developers to small businesses and enterprise AI teams.
According to the company, a single UGen B70 supports most models below 27 billion parameters and can provide a context window of up to 128K tokens with Llama 3.1 8B.
Software support for local AI workloads
ASUS says the UGen B70 is supported by an AI software ecosystem that includes Intel LLM Scaler, vLLM XPU, llama.cpp through SYCL and Vulkan, Ollama, Intel OpenVINO and oneAPI, PyTorch XPU, and ComfyUI.
The company lists verified models including Llama 3.1 and 3.2, Qwen 3.x, DeepSeek-R1 Distill, Mistral and Mixtral, OpenAI gpt-oss, Whisper-large-v3, and InternVL3.5. ASUS claims that a dual-card configuration has been demonstrated running 80B-class models at 43.4 tokens per second.
Target users and multi-card deployment
ASUS is targeting three main groups with the UGen B70. For prosumers and independent developers, the company positions a single 32GB card as an option for running larger models locally than mainstream 16GB graphics cards can typically support.
For small businesses and creative studios, ASUS proposes dual-card 64GB configurations to support larger models while keeping sensitive data on-premises.
For enterprise departmental AI teams, the company positions four-card 128GB configurations as an alternative to ongoing cloud API usage, particularly in sectors such as law, medicine, government, and finance where data compliance may be important.
ASUS describes this scalable local deployment model as part of its broader “Agent Computer” concept, referring to systems designed to run AI agents locally rather than relying exclusively on cloud infrastructure.
Availability and pricing
The ASUS UGen Intel Arc Pro B70 32GB will be available worldwide through authorized ASUS resellers and distributors. Pricing will be announced closer to availability.
