qwen38-flash-next
在本地一键运行 Qwen3.8-Flash-Next（125B MoE，6B 激活）UD-IQ1_S 量化版。
执行后自动完成：查找或安装 llama.cpp 运行时（qwen4exp 架构需要未合入的 PR #27742 构建）→ 查找或下载模型（默认走 hf-mirror.com，国内网络友好）→ 启动本地服务 → 进入交互式对话。已存在的模型和运行时会被直接复用，不重复下载/编译。
安装
pip install qwen38-flash-next # PyPI
# 或本地源码
uv tool install ./qwen38-flash-next # 推荐，隔离环境
使用
qwen38-flash # 交互式对话
qwen38-flash --say "你好" # 单发一句（脚本/测试用）
qwen38-flash --download-only # 只下载模型
qwen38-flash --ms # 从 ModelScope 下载（hf-mirror 不通时用）
首次运行需要约 74GB 磁盘与下载时间；48GB 内存机器默认 -ngl 0 纯 CPU 运行（Metal 会因超大嵌入层 OOM）。
交互命令：/exit 退出，/clear 清空上下文，/save 文件 保存对话。
常用选项
| 参数 / 环境变量 | 说明 |
|---|---|
--model PATH / QWEN38_MODEL |
指向模型首分片 ...00001-of-00003.gguf，跳过查找与下载 |
--server-bin PATH / QWEN38_LLAMA_SERVER |
指定支持 qwen4exp 的 llama-server 二进制 |
--port 8080 / --ctx 8192 / --ngl 0 |
服务端口 / 上下文 / GPU 层数 |
HF_ENDPOINT |
模型下载源，默认 https://hf-mirror.com |
模型与运行时路径在首次就绪后写入 ~/.qwen38-flash/config.json，后续启动直接复用。
查找顺序（逐级复用）
模型：--model/环境变量 → 配置文件 → 常见位置（HF 缓存、ModelScope 缓存、~/models、本仓库工作区）→ 自动下载。
运行时：--server-bin/环境变量 → 配置文件 → PATH 中的 llama-server（会校验能否加载模型）→ 自动克隆 PR #27742 源码编译（需 git + cmake + Xcode CLT）。
1 maintainer
Release files for qwen38-flash-next 0.1.4
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| qwen38_flash_next-0.1.4.tar.gz | 10.5 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| qwen38_flash_next-0.1.4-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 22.6 kB
Release files / qwen38_flash_next-0.1.4.tar.gz
| Download URL | qwen38_flash_next-0.1.4.tar.gz |
|---|---|
| Size | 10.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
9bd7cffc2309955f3a7eed86c61fe21c9b5c7edc184877378a38334d8d9c652e
|
|
BLAKE2b-256 checksum How to use checksums |
d7fdb2149817b598c508257ea3745c5ce76e78d133c35a57cf3d542432e477eb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.12.5 {"installer":{"name":"uv","version":"0.12.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / qwen38_flash_next-0.1.4-py3-none-any.whl
| Download URL | qwen38_flash_next-0.1.4-py3-none-any.whl |
|---|---|
| Size | 12.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
177a467f2d789ced74ce593eadae28b8fe33b29deeac0f2bbf59dd215390b233
|
|
BLAKE2b-256 checksum How to use checksums |
f0ee75f946995120bcaa7354e8d241e55804637ccfdd6ac85219a30ba946c336
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.12.5 {"installer":{"name":"uv","version":"0.12.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
