llama.cpp v0.5.0 发布：后端性能提升与多模型支持

AI 摘要

llama.cpp 发布 v0.5.0，重点提升后端性能与正确性、扩展模型覆盖并增强服务端/路由稳定性。新增 HRM-Text（DFM Mimir 1B）、MiMo-V2.6 与 HunyuanOCR 转换支持，升级 ggml 0.25.0，并加入多地址 HTTP 绑定、函数调用图像输出及多项聊天解析器/UI 修复。

正文节选

## Overview This release focuses on backend performance and correctness, broader model coverage, and more robust server/router operation. It adds HRM-Text (DFM Mimir 1B) support, MiMo-V2.6 and HunyuanOCR conversion support, ggml 0.25.0 backend improvements, multi-address HTTP binding, image outputs from function calls, and several chat parser/UI fixes. ### Highlights - Accelerate CUDA `conv2d` with implicit GEMM ([#29135](https://github.com/ggml-org/llama.cpp/pull/29135)) - Add Metal M

发布时间：2026-09-24 04:50

抓取时间：2026-09-24 05:05

来源机构：ggml-org