b580

Bonsai 2 27B ternary on Intel Arc B580: SYCL + XMX, 128K context

Qwen Image-2.1 on Intel Arc B580

Run LLMs on Intel Arc (B60, B580, B570, B50, B70, A770) with llama.cpp Vulkan: benchmarks, tuning, agent reliability evals, raw data.

Run Krea 2 (12B DiT) on an Intel Arc B580 in Docker — working PyTorch-XPU/ComfyUI setup for Battlemage on kernel 7.x, plus the fix for the Level-Zero abort, benchmarks, and troubleshooting

llama.cpp SYCL patches and tuning for Intel Arc B580 12 GB: 35B-A3B MoE at 47-50 t/s with 160-200K context, and ternary Bonsai 2 27B at 196K - on a single 12 GB card

Fork of llama.cpp with Intel Arc / SYCL build configs and A770 tuning. Pre-tested Docker + native Arch Linux setup. One-command install script. Solves: SYCL kernel module miss, AOT cache, optimal quant picks for 16GB.