Ollama 发布 v0.40.16：OpenAI 接口与 MLX 优化

AI 摘要

Ollama 发布 v0.40.16 版本，包含多项更新：OpenAI 兼容接口支持在响应请求中转发截断选项，未安装模型时 /v1/models 返回空数组；llama.cpp 升级至 b12177；Windows 平台新增 CUDA 13.5 与 ARM64 的 SM 103 支持；MLX 优化 Qwen3.8 Flash Next 的内存效率并新增 ripple 预览支持；同时修复 Claude Code 路由与已退役模型报错问题。

正文节选

## What's Changed * openai: forward truncation option in responses requests by @marlonferreira1800-hue in https://github.com/ollama/ollama/pull/19501 * openai: return an empty array for /v1/models when no models are installed by @liangliang205 in https://github.com/ollama/ollama/pull/19539 * llama.cpp: version bump b12177 by @dhiltgen in https://github.com/ollama/ollama/pull/19576 * README: add Quock to Mobile community integrations by @matteocelani in https://github.com/ollama/ollama/pull/1

发布时间：2026-11-10 02:21

抓取时间：2026-11-10 02:53

来源机构：Ollama