GPT4All review
Discontinued: Nomic's free, open-source desktop app for running LLMs locally, unmaintained since its last release (v3.10.0) in February 2025.
By Noahlast updated pricing verified
GPT4All screenshots & video
GPT4All in 20 seconds
Best for: existing users keeping an offline GPT4All setup running while they migrate to a maintained local LLM runner
Pricing: No longer developed; the final v3.10.0 build (February 2025) remains a free, MIT-licensed download with no paid tier.
- GPT4All's last release, v3.10.0, shipped on February 24, 2025, and its GitHub repo has had no commits since May 27, 2025.
- On December 19, 2025, Nomic said on GitHub issue #3638 that it was working to put GPT4All into "non-maintained maintenance mode".
- GPT4All is MIT-licensed, and its GitHub repo has more than 77,000 stars.
- GPT4All ran models locally on Windows, macOS and Linux, with LocalDocs document chat and an OpenAI-compatible local server on port 4891.
- The installers on nomic.ai/gpt4all still download the final v3.10.0 build.
Limitations: No longer developed or supported; no release since February 2025.; The unauthenticated local API server never received the DNS-rebinding fix in pull request #3587.; Cannot load many model architectures released after early 2025.
Top alternatives: LM Studio, Jan, Ollama, AnythingLLM, Msty, Open WebUI — full list
What is GPT4All?
GPT4All is a desktop application from Nomic that downloads quantized GGUF language models from Hugging Face and runs them on the user's own CPU or GPU through a llama.cpp-based backend, with no account, API key or internet connection needed once a model is on disk. Its best-known feature, LocalDocs, indexes a folder of local files with Nomic's on-device embedding models so chats can quote and cite those documents, and an optional local server exposes an OpenAI-compatible API on port 4891. Nomic released GPT4All in March 2023 and shipped a redesigned v3.0 in July 2024, but development stopped after v3.10.0 in February 2025: the repo has had no commits since May 2025, and in December 2025 Nomic said it was moving the app to unmaintained status. The final build still installs and runs, yet its backend gets no updates, its model catalog stops at early-2025 models, and a security fix for its unauthenticated local API server was never merged.
GPT4All is Nomic's free, MIT-licensed desktop app for chatting with open-weight large language models entirely offline on Windows, macOS and Linux, and it is no longer maintained. Its last release, v3.10.0, shipped on February 24, 2025, the last commit to its GitHub repo landed on May 27, 2025, and on December 19, 2025 Nomic's Andriy Mulyar wrote on the project's issue tracker (issue #3638) that the company was "working to formally put the app in non-maintained maintenance mode"; nomic.ai itself now sells AI agents for architecture, engineering and construction firms. The installers on nomic.ai/gpt4all still download that final build. Because GPT4All runs locally, chats, downloaded models and LocalDocs indexes stay on users' own machines; nothing was deleted. Former users have moved to actively developed local runners such as LM Studio, Jan, Ollama and AnythingLLM.
Standout feature: LocalDocs was GPT4All's signature feature: point it at a folder and it indexed the files with Nomic's on-device embedding models, so fully offline chats could answer from and cite your own documents.
Biggest weakness: GPT4All is no longer developed: no release since v3.10.0 (February 24, 2025), no commits since May 27, 2025, and a pull request that blocks DNS-rebinding attacks on its unauthenticated local API server (#3587) has sat unmerged since May 2025, so the app gets neither security fixes nor support for newer model architectures.
GPT4All: pros and cons
Pros
- Ran open-weight models fully offline on ordinary laptops, with no account, API key or dedicated GPU required.
- LocalDocs indexed local folders with on-device Nomic embeddings and showed which files an answer came from.
- MIT-licensed, with native installers for Windows (x64 and ARM), macOS 12.6+ and Ubuntu, plus a Python SDK (pip install gpt4all).
- An optional OpenAI-compatible local API server on port 4891 let other tools use the loaded model.
Cons
- No longer maintained: the last release was v3.10.0 on February 24, 2025, and the repo has had no commits since May 27, 2025.
- The local API server has no authentication, and the fix for DNS-rebinding attacks (pull request #3587) was never merged.
- The built-in model catalog stops at early-2025 models such as the DeepSeek-R1 distills, and architectures added to llama.cpp later are not supported.
- The Python SDK (gpt4all on PyPI) has not had a release since version 2.8.2 in August 2024.
GPT4All pricing
Free
What reviewers say
No third-party review scores on file yet.