在 AMD Strix Halo 上本地跑 Qwen3.8-Flash-Next 和 NPU 小模型，用同一端口同时提供 OpenAI 与 Anthropic 兼容 API，让本地推理像调前沿云服务一样顺手。
把普通 PC、Mac 或 Linux 机器变成私有 AI 服务器，集成 llama.cpp、ComfyUI、Open WebUI、n8n 等组件，方便本地跑大模型、RAG 和 Agent 工作流。
在单台桌面机器上跑 100GB 级大模型的开源推理引擎，支持 AMD Strix Halo、NVIDIA RTX 和 Apple Mac，兼容 OpenAI API，基于 ExLlamaV3 构建。
…00 on (CUDA, Linux and Windows) and AMD (HIP, the Strix and Gorgon Halo included). Not on Mac or Pascal cards. About 256 GB of VRAM and RAM together keeps every expert in
…epSeek reads prompts (prefill) much faster on AMD Strix / Gorgon Halo and faster on multi-GPU NVIDIA PCs, Qwen-3.8-Flash-Next itself as unsloth's UD-IQ2 XS GGUF, and four
…m/kaito-project/aikit/pull/838) - 🟠 AMD ROCm and Strix Halo support. Build hardware-accelerated inference images for supported AMD GPUs. Thanks [@keithmattix](https://gi
…ct-07e2a21e3cca-harness-19d43e6f9f25-hosts-tower2-strix-halo-spark-m5-mbp-windows-laptop-strixy - Passed: regressions 16/16 , zero-prereq bootstrap 6/6 , fresh install, v
…fecycle, and distro lab validation across tower2, Strix Halo, Spark, and M5 MacBook Pro. - The new dream-proxy-owner-card-readiness-1474 regression fixture passed on Stri
…ties, lifecycle, and UI validation across tower2, Strix Halo, Spark, and M5 MacBook Pro. - Full-model capability probes passed on all 4 enabled hosts, including chat, sea