← Runners

GPT4AllGPT4All

Nomic AI

开源Open source MITMIT 桌面应用Desktop application CPU优先CPU-first llama.cpp系llama.cpp-based partial lifecycle: maintenance 核验Checked 2026-10-30

GPT4All 是「人人设备上跑本地模型」的桌面工具（Nomic AI，MIT）：CPU 优先、桌面开箱、商用无附加条件。诚实标注：仓库最近推送 2025-05-27（API 快照）——热度在历史里，维护节奏要看当下。

GPT4All is the founding standard-bearer of 'local models on everyone's device' (Nomic AI, MIT): CPU-first, desktop out of the box, commercial use with no strings. Honest note: last push 2025-05-27 (API snapshot) — the stars are history, check the current maintenance cadence.

适合与不适合When to use it, when not to

适合：无独显/低配机器跑小模型、要 MIT 干净许可的桌面方案。不适合：要最新模型开箱支持；要服务化或多模型调度。

Fits: small models on low-spec/CPU machines, a clean MIT desktop option. Not a fit: day-one support for new models; serving or multi-model scheduling.

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　Nomic AI（美国）。Nomic AI (US).
国内使用Direct access: 大陆可直连direct OK　nomic.ai 直连实测 200（2026-10-21）。nomic.ai measured 200 direct (2026-10-21).
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: 官方模型目录（GPT4All 格式 .ggjt/.gguf 线）+ 本地导入；模型目录可用性随维护节奏变化—— 未逐条清点。
Official model catalog (the GPT4All-format .ggjt/.gguf line) + local import; catalog availability shifts with the maintenance cadence — not inventoried item by item.
运行位置Where it runs: 桌面应用（Windows/macOS/Linux）；基于 llama.cpp 系。GitHub 77,349★，最近推送 2025-05-27（GitHub API 核验 2026-10-30——发版放缓是事实，如实记）。
Desktop app (Windows/macOS/Linux); built on the llama.cpp line. GitHub 77,349★, last push 2025-05-27 (GitHub API verified 2026-10-30 — the release slowdown is a fact, recorded as such).
本地文件Local files: 模型与对话默认落本机；官方口径是本地运行，未见官方云线（与 Ollama 的云端线不同）。
Models and conversations land on your machine by default; the official stance is local running, no official cloud line seen (unlike Ollama's cloud line).
关机后的任务Tasks after shutdown: 内置本地 API server 线可供外部客户端连接；该线当前状态与默认开关未逐条核验。
A built-in local API server line can be connected by external clients; its current status and default switch not verified item by item.
工具与扩展Tools & extensions: 应用内聊天为主轴；agent/工具调用生态未见官方主线，第三方扩展面也未逐条核验。
In-app chat is the main axis; no official main line seen for an agent/tool-calling ecosystem, and the third-party extension surface is not verified item by item.
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能力未见官方声明（未核验到≠没有）。
Session context management is in-app; cross-session memory and long-term memory capability — 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、无云同步权限面（未见官方云线声明）。
Desktop-app permissions follow the OS grant; no account system, no cloud-sync permission surface (no official cloud-line statement seen).
适合什么任务Best for: 适合：低配机器（CPU 优先）跑小模型、要 MIT 无附加条件的桌面方案。不适合：要新模型开箱（维护节奏放缓——看仓库最近推送）；要服务化部署。
Fits: running small models on low-spec machines (CPU-first), wanting a MIT no-strings desktop solution. Does not fit: wanting new models out of the box (maintenance cadence has slowed — check the repo's last push); wanting service deployment.

头号误解Top misconception

把 77k★ 的热度当现行维护度——最近推送 2025-05-27（API 快照），选型前看仓库节奏Reading 77k stars as current maintenance — last push 2025-05-27 (API snapshot); check the cadence before choosing
把「commercial use」承诺延伸到模型权重——权重许可跟着模型走Extending the 'commercial use' promise to model weights — weight licenses follow the model
期待新模型跟进——放缓节奏下新架构支持不保证Expecting fast support for new architectures — under a slower cadence it is not guaranteed

未知项清单Known unknowns

官方模型目录当前可用清单（未逐条清点）
维护放缓的官方说明（未见——只记仓库事实，不猜原因）
内置 API server 的当前状态（未逐条核验）

The current usable list of the official model catalog (not inventoried item by item)
An official explanation of the maintenance slowdown (none seen — only repo facts recorded, no guessing at causes)
The current status of the built-in API server (not verified item by item)

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
repo	nomic-ai/gpt4all · 仓库（MIT；77,349★，最近推送 2025-05-27，GitHub API 核验 2026-10-30）nomic-ai/gpt4all · repo (MIT; 77,349★, last push 2025-05-27, GitHub API verified 2026-10-30)	https://github.com/nomic-ai/gpt4all
docs	GPT4All · 官网产品页GPT4All · official product page	https://www.nomic.ai/gpt4all
docs	GPT4All · 官方文档GPT4All · official docs	https://docs.gpt4all.io
changelog	nomic-ai/gpt4all · Releases（changelog）nomic-ai/gpt4all · Releases (changelog)	https://github.com/nomic-ai/gpt4all/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).