AI NewsProductivityAnnouncement
Parallio is an open source desktop app that puts several AI models in one group chat, so a user can ask the same question of Claude, GPT, Gemini and DeepSeek side by side and compare the answers in one window
A developer using the handle richard1208 released Parallio, an MIT-licensed Electron desktop app and Chrome extension that lets a user pose one question to several AI models at once and compare their answers in the same window. The repository was created on 24 September 2026 and reached 204 stars and 10 forks in six days.
Image: GitHub / richard1208
Why it mattersA researcher, product manager or writer comparing model answers today does it by copying the same question into four browser tabs, and Parallio removes the copy-paste step and the four separate cost trackers that came with it.
A researcher or writer comparing what four AI models say about the same question does it today by opening four browser tabs, pasting the same prompt into each, and reading four answers in four different places. On 24 September 2026 a developer using the handle richard1208 published Parallio, an MIT-licensed Electron desktop app and Chrome extension that puts several models in one group chat, and it reached 204 stars and 10 forks on GitHub in six days.
What it does
The desktop side is one window with a shared timeline. A user types a question once, attaches up to ten files or 100 MB, and picks which of the connected models should answer. Each model responds in the same window, and the cost of each answer is shown next to it. The developer's README says the desktop app connects through the user's own OpenRouter key and covers GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi and several others.
Chats and messages are stored in a local SQLite file, not a server the developer runs. The final report from a session exports to DOCX, PDF, XLSX or PPTX. The interface is available in English and Chinese.
The Chrome extension is the other half of the project and works on model websites the user already has open, sending the same question to six providers on the Chinese market: Doubao, DeepSeek, Qwen, Zhipu, Kimi and Yuanbao. The two lists do not overlap fully; the Chinese-only providers on the extension are not on the desktop, and the Western providers on the desktop are not on the extension.
The traction and the version
The repository at github.com/richard1208/parallio was created on 24 September 2026 and has 204 stars and 10 forks as of publication. There are no open issues on the tracker. The project is at version 0.1.12, which the developer labels experimental. The desktop app requires Node.js 22.13 or later to build; the release track packages Windows and Mac binaries but the Windows build is not code-signed.
That growth rate, 204 stars in six days, is real for a niche tool. It is not the same category as an open source alternative to a paid product with a wide audience, and the developer says as much: the README warns that the desktop side was tested against a simulated OpenRouter and that the extension needs real-world validation across all six supported websites.
Why the pattern is worth watching
Comparing model answers side by side is what a working researcher, product manager or content writer does often enough that they have their own way of doing it, usually badly. The paid tools for this, such as OpenRouter's own dashboard and Poe, are aimed at chat with one model at a time, and the workaround is four browser tabs. A local group chat that runs against the user's own keys, keeps the transcript in a file on the user's computer, and shows the cost of each model in the same view, is the small tool that fills the gap. Parallio may or may not be the one that people settle on, and version 0.1.12 says as much, but the shape of the tool is the news.
Source
Primary source: Parallio repository on GitHub, richard1208/parallio, created 24 September 2026, last push 30 September 2026.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
