Locally Uncensored vs Jan.ai

Jan.ai has earned a reputation as one of the most polished, ChatGPT-like interfaces among local AI apps. Built on Tauri with a clean design, it supports both local models and cloud APIs from providers like OpenAI, Anthropic, Mistral and Groq. For users who want a single chat interface that bridges local and cloud AI, Jan is an appealing option.

Locally Uncensored takes a different approach. Rather than trying to be a universal chat client, it aims to be a complete local AI studio. It combines LLM chat with image and video generation, which Jan's feature list does not include, and adds a coding agent and an agent with tools. If you are exploring the best local AI apps in 2026, these two represent very different philosophies about what a local AI app should be.

Both apps can run on an engine you pick yourself, and if that choice is still open, LU Labs has an Ollama vs LM Studio comparison.

Jan runs on Windows, macOS, and Linux. Locally Uncensored runs on Windows and Linux, and since version 3.0.6 on Macs with Apple Silicon and macOS 14 or newer, as a beta. Both let you chat with local models on your own hardware. This comparison breaks down where each app does better and where it falls short.

Feature Comparison

Where Locally Uncensored Wins

• Image and video generation: the biggest difference. On Windows and Linux, Locally Uncensored installs and runs ComfyUI for you and offers image models such as FLUX, Juggernaut XL and Qwen-Image 2.1 and video models such as Wan, HunyuanVideo and LTX. On a Mac, Create runs on the Mac itself. Jan's feature list includes no image or video generation.
• AGPL-3.0 license: Locally Uncensored is open source under AGPL-3.0, which keeps changes to the code open. Jan is open source too, under the permissive Apache 2.0 license.
• Custom personas: 25 built-in AI personas let you switch between different personalities instantly without writing system prompts. In Jan you build custom assistants yourself.
• Answers directly, abliterated models available: pick a model with the refusal behaviour removed from the weights, or a mainstream one, in the same picker. Nothing sits between you and the model on your own machine.
• Coding agent: it builds a map of the repo, edits only the lines you asked for, shows the diff before it applies anything, and runs your tests. Jan's coding agent, Jan Agent, is a separate terminal app in preview.
• Agent tools and MCP: web search and fetch, file operations, a shell, code execution, screenshots and image and video generation are built in, and your own MCP servers add more. Long jobs can run as background agents.
• Multi-provider support: local engines and cloud providers with your own keys in one picker, which closes most of the gap to Jan's cloud API support.

Version 3.0.6 also brings document chat with a local index, phone access over your LAN or a tunnel, side by side model compare and a local benchmark, and two beta features: Voice, a spoken conversation run on your own machine, and computer control, which is off until you switch it on.

What Locally Uncensored 3.0.7 adds

Version 3.0.7 adds these points that matter next to Jan:

• Duplicate and branch: Duplicate chat copies a conversation with its persona and settings, Branch from here starts a new chat from any message, and Regenerate keeps the earlier answer as a version you can switch back to. These have run in automated tests only.
• Several local servers at once: KoboldCpp, LM Studio, a llama.cpp server or any other OpenAI compatible backend each get their own row in Settings, AI Backends, and their models sit in the picker under that name. Adding one no longer pushes another out, and a group chat can mix models from several of them. The context window comes from each server itself. So far this has run in automated tests only, not yet in the app with two real servers.
• Reasoning effort for local models: the Low, Medium, High and Max button next to Think now works on the LU Engine, on LM Studio 0.4.8 and later, on KoboldCpp (Low to High) and on a llama.cpp server whose chat template reads it. It is one setting for all models, and it has run in automated tests only, not against a real LM Studio, KoboldCpp or llama.cpp server. A separate think budget in tokens works with the LU Engine and with Anthropic.
• Any GGUF file on your drive: the Model Manager links a GGUF file from anywhere on the same drive into the LU Engine as a chat model, without copying it.
• ChatGPT export as real chats: the .zip of a ChatGPT data export turns into separate chats with their titles and dates, and a second import adds nothing twice. Tried in the app with an export built in ChatGPT's format; a real export from ChatGPT has not been tried yet.
• Character cards: a card as PNG or JSON (versions 1, 2 and 3) becomes a persona with the card picture. A new chat opens with the card's greeting, alternate greetings switch with arrows, and a card persona exports back as a version 2 JSON file. Checked in the app with Holo and Qwen models, not with Gemma.
• More of Create on a Mac: Talking Character, Motion Control, Music, Enhance Image, Remove Background, masks in Edit and character training now run on the Mac itself. They ran with the real engine on one Mac with 48 GB; Macs with less memory have not been tried, and the Mac parts have not been clicked through in a built app yet.

The Settings, Local API page still reaches only the LU Engine, Ollama and LM Studio. Our check of 3.0.7 in the real app ran on one Windows machine.

Where Jan.ai Wins

• Cloud API support: Jan connects to OpenAI, Anthropic, Mistral, Groq, MiniMax and other cloud providers alongside local models. Locally Uncensored now takes OpenAI, Anthropic, OpenRouter, Groq, Together, DeepSeek, Mistral and any OpenAI compatible address, so this gap is much smaller than it was.
• Polished chat interface: Jan has one of the cleanest, most intuitive chat UIs among local AI apps. It genuinely feels like a local version of ChatGPT. The conversation management, threading, and overall design are excellent.
• MCP with inline approval: Jan connects MCP servers and shows each tool call with its arguments in the chat before you accept or deny it.
• HuggingFace integration: download models from Hugging Face in Jan's Hub, or import GGUF files you already have. Locally Uncensored has a GGUF search on Hugging Face in its model manager next to its own curated catalog.
• Longer track record on the Mac: Jan has run on macOS for a long time. Locally Uncensored has had a Mac build only since 3.0.6, for Apple Silicon and as a beta.

Architecture Differences

The two apps are closer in architecture than their feature lists suggest. Jan is built with Tauri and ships llama.cpp inside the app, updated together with it, plus MLX on a Mac. Since version 0.8.0 one router process loads and unloads models on demand.

Locally Uncensored uses Tauri (Rust backend with native system webview). For text it ships its own LU Engine and also works with an engine you already run: it detects twelve of them, including Ollama, LM Studio, Jan and llama.cpp. Images and video run through ComfyUI on Windows and Linux. A first-run wizard finds what is already installed and installs what is missing with one click, ComfyUI included.

For users who already have Ollama or LM Studio, Locally Uncensored adds no second inference engine and can link the models those tools already store. For users who start from nothing, both apps now come with an engine of their own.

What this page says about Jan comes from its official website, documentation, changelog and GitHub repository, as of October 2026.

More comparisons

• Alternatives and comparisons, every head to head on one page
• Locally Uncensored vs Jan, the short version
• What Locally Uncensored is, what runs on your own machine and what runs on LU Cloud