JanJan
Jan 是「开源版 ChatGPT 桌面替代」的长期主义者:100% 离线跑本地模型,Apache-2.0 主体加一句署名请求。 诚实标注:GitHub API 标 NOASSERTION 是机器判不了附加署名句——读 LICENSE 原文才准;官网直连 000(网络受限非死亡)。
Jan is the long-standing open-source 'offline ChatGPT replacement': 100% offline local models, Apache-2.0 core plus an attribution request. Honest notes: the GitHub API's NOASSERTION flag comes from that extra sentence — read the LICENSE itself; the site measured 000 direct (network-limited, not dead).
适合与不适合When to use it, when not to
适合:要开源可审计的桌面离线聊天、要内置 OpenAI 兼容本地端点。 不适合:服务化部署;要最顺滑的模型获取体验。
Fits: auditable open-source offline desktop chat, built-in OpenAI-compatible local endpoint. Not a fit: serving deployments; the smoothest model-acquisition experience.
中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-21
- 出品属地Origin
- 国外出品foreign-made Menlo Research(美国/新加坡背景的研究机构)。Menlo Research (US/Singapore-background research org).
- 国内使用Direct access
- 直连不可达direct blocked www.jan.ai 直连实测 000(网络受限,非死亡判定;经代理可达 200)。www.jan.ai measured 000 direct (network-limited, not a death verdict; 200 via proxy).
- 备案状态CAC filing
- 未见于备案名单not in filing list:未见于网信办公示名单(2026-10-21 检索);该产品未面向大陆公众提供服务,备案不适用。not in the CAC public registry (searched 2026-10-21); the product does not serve the mainland public, so filing does not apply.
备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准(按批次滚动更新,本站只抄编号不解读);「大陆可直连」是本站从大陆网络直连实测的结果,不是厂商承诺。GitHub 类站点当日实测有波动,口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.
固定坐标系Fixed coordinate system
8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.
- 模型与开放条件Model & openness
官方定位「100% offline on your computer」(仓库描述直抓所见);GGUF 本地模型为主, 模型清单随版本滚动,未逐条清点。
Official positioning "100% offline on your computer" (as seen in the repo description, direct-fetched); GGUF local models are the mainstay, the model list rolls with versions, not inventoried item by item.
- 运行位置Where it runs
桌面应用(Jan 自研运行时 jan-core/jan-engine 线);GitHub 45,158★(API 核验 2026-10-21)。 仓库现归 Menlo Research 名下(janhq/jan;menloresearch/jan 为旧路径 301 跳转,直抓核验)。
Desktop app (Jan's own runtime: the jan-core/jan-engine line); GitHub 45,158★ (API verified 2026-10-21). The repo now sits under Menlo Research (janhq/jan; menloresearch/jan is the old path with a 301 redirect, direct-fetch verified).
- 本地文件Local files
对话与模型全落本机(官方「100% offline on your computer」口径,仓库描述直抓所见)。
Conversations and models all land on your machine (the official "100% offline on your computer" stance, as seen in the repo description, direct-fetched).
- 关机后的任务Tasks after shutdown
应用内置本地 API server(OpenAI 兼容)可常开;细节未逐条核验。
A built-in local API server (OpenAI-compatible) can stay on; details not verified item by item.
- 工具与扩展Tools & extensions
插件/扩展生态在官方文档线内;MCP 支持状态未逐条核验,第三方扩展质量未清点。
A plugin/extension ecosystem is within the official docs line; MCP support status not verified item by item, third-party extension quality not inventoried.
- 上下文与记忆Context & memory
会话上下文管理在应用内;跨会话记忆与长期记忆未见官方声明(未核验到≠没有)。
Session context management is in-app; cross-session memory and long-term memory — no official statement seen (not-verified ≠ absent).
- 权限与限制Permissions & limits
桌面应用权限随操作系统授权;无账号体系,离线定位下未见云同步权限面。
Desktop-app permissions follow the OS grant; no account system, and under the offline positioning no cloud-sync permission surface seen.
- 适合什么任务Best for
适合:要开源的桌面 GUI 跑离线模型(对比闭源的 LM Studio)。 不适合:要服务化多模型部署;要开箱即用最新模型(模型获取体验弱于 Ollama)。
Fits: wanting an open-source desktop GUI for offline local models (vs closed-source LM Studio). Does not fit: service-style multi-model deployment; wanting latest models out of the box (the model-acquisition experience is weaker than Ollama's).
头号误解Top misconception
- 把 NOASSERTION 当「许可不明」——读 LICENSE 原文是 Apache2.0+署名请求,GitHub 标记只是机器判不了附加句Reading NOASSERTION as 'license unclear' — the LICENSE is Apache-2.0 plus an attribution request; the API flag just cannot parse the extra sentence
- 用旧仓名 menloresearch/jan 找仓库——现路径 janhq/jan(旧路径 301 跳转仍可达)Using the old path menloresearch/jan — the repo now lives at janhq/jan (old path 301s)
- 把「开源」与「模型免费」混谈——权重许可跟着模型走Mixing up 'open source' with 'models are free' — weight licenses follow the model
未知项清单Known unknowns
- jan-core/jan-engine 运行时的模型格式支持矩阵(未逐条核验)
- MCP 支持状态(未逐条核验)
- 署名要求的法律约束力边界(Apache 附加句是请求还是条件——本站不做法律判断)
- The model-format support matrix of the jan-core/jan-engine runtime (not verified item by item)
- MCP support status (not verified item by item)
- The legal-binding boundary of the attribution request (is the Apache extra sentence a request or a condition — this site makes no legal judgment)
证据来源Evidence sources
判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-30,内容更新日content updated on 2026-10-21。
| 类型Type | 名称Name | 链接Link |
|---|---|---|
| docs | janhq/jan · LICENSE(Apache2.0+署名请求,2026-10-21 直抓核验)janhq/jan · LICENSE (Apache2.0 + attribution request, direct-fetched and verified 2026-10-21) | https://github.com/janhq/jan/blob/main/LICENSE |
| repo | janhq/jan · 仓库(45,158★,最近推送 2026-10-30,GitHub API 核验)janhq/jan · repo (45,158★, last push 2026-10-30, GitHub API verified) | https://github.com/janhq/jan |
| docs | Jan · 官网Jan · official site | https://www.jan.ai |
| changelog | janhq/jan · Releases(changelog)janhq/jan · Releases (changelog) | https://github.com/janhq/jan/releases |
实测记录Measured runs
本档未跑吞吐与显存实测;结论只覆盖运行时能力与许可。
This site has not completed measured runs. The closest test protocol is the code-side protocols/framework/_protocol.md (the local-endpoint and credential-boundary sections are reusable; no runner-specific protocol exists on this site yet).
相关条目Related entries
- LM StudioLM Studio同为桌面 GUI:LM Studio 闭源但 agent 叙事激进,Jan 开源但节奏更稳。Both desktop GUIs: LM Studio is closed-source but aggressive on the agent narrative, Jan is open-source with a steadier cadence.
- OllamaOllama模型获取与 CLI 线 Ollama 更强;Jan 的价值在开源桌面壳。Ollama is stronger on model acquisition and the CLI line; Jan's value is the open-source desktop shell.
- OpenAI Agents SDKOpenAI Agents SDK另一层 · 框架与运行时Other layer · Frameworks & runtimes
本页由This page is generated from the ai-handbook 数据层生成(CC BY 4.0)。data layer (CC BY 4.0).
方法论与坐标系定义见仓库内For the methodology and coordinate definitions, see METHODOLOGY.md.