Lemonade v10.7 release and project organization update
English summary
Lemonade v10.7 introduces local omni-modal chat supporting image generation and editing by combining multiple backends and models; its LMX-Omni virtual models are now compatible with Open WebUI and other OpenAI clients. The release adds a lemonade bench CLI tool to collect standardized LLM performance data across llama.cpp, FastFlowLM, and vLLM. Cross-vendor support expands with CUDA backends for llama.cpp and stable-diffusion.cpp and a Vulkan backend for sd-cpp, enabling GPU acceleration on AMD, Apple Silicon, Nvidia, and Intel systems. The project is now organized into six working groups, four led by non-AMD contributors, and this release involved 19 contributors.
Chinese summary
Lemonade v10.7 推出本机全模态聊天，通过组合多个后端和模型支持图像生成与编辑，其 LMX-Omni 虚拟模型现已兼容 Open WebUI 及其他 OpenAI 客户端。该版本新增 lemonade bench 命令行工具，可在 llama.cpp、FastFlowLM 和 vLLM 之间收集标准化的 LLM 性能数据。跨厂商支持得到扩展，为 llama.cpp 和 stable-diffusion.cpp 添加 CUDA 后端，为 sd-cpp 添加 Vulkan 后端，实现在 AMD、Apple Silicon、Nvidia 和 Intel 系统上的 GPU 加速。项目现已划分为六个工作组，其中四个由非 AMD 贡献者领导，本版共有 19 位贡献者参与。
Key points
Local omni-modal chat combines backends and models for image generation/editing, with LMX-Omni virtual models now compatible with Open WebUI and OpenAI clients.
本机全模态聊天通过组合后端与模型实现图像生成/编辑，LMX-Omni 虚拟模型现已兼容 Open WebUI 和 OpenAI 客户端。
New lemonade bench CLI tool enables apples-to-apples LLM performance comparisons across llama.cpp, FastFlowLM, and vLLM.
新增 lemonade bench 命令行工具，可在 llama.cpp、FastFlowLM 和 vLLM 之间进行同类 LLM 性能对比。
Cross-vendor GPU support adds CUDA backends for llama.cpp and stable-diffusion.cpp, plus Vulkan for sd-cpp, accelerating on AMD, Apple Silicon, Nvidia, and Intel.
跨厂商 GPU 支持新增 llama.cpp 和 stable-diffusion.cpp 的 CUDA 后端，以及 sd-cpp 的 Vulkan 后端，加速覆盖 AMD、Apple Silicon、Nvidia 和 Intel。
Project restructured into six working groups, with four led by non-AMD contributors; v10.7 had 19 contributors.
项目重组为六个工作组，其中四个由非 AMD 贡献者领导；v10.7 共有 19 位贡献者参与。