LumiChats

LumiChats is a dual-product AI platform built by students for students and developers. The web platform provides access to 40+ frontier AI models — including Claude, Gemini, GPT-4, DeepSeek, and Qwen — in a single browser tab, while the offline Windows desktop app runs GGUF-format models entirely on-device with no internet, no GPU, and no data collection. The platform launched publicly in January 2026 via Product Hunt and, according to the company's own About page, serves 30,000+ total users across 80+ countries.

What It Is

LumiChats sits at the intersection of a multi-model AI aggregator and a study tool. Rather than requiring separate subscriptions for each AI provider, it brings every major frontier model into one interface. The web app adds Study Mode — a page-locked RAG feature that constrains AI answers to specific pages of an uploaded PDF, citing exact page numbers — alongside Agent Mode, Quiz Hub, and persistent memory. The offline app (LumiChats Offline, v1.1) is built on top of GPT4All by Nomic AI under the MIT license and ships with six fine-tuned LumiChats models ranging from 1B to 4B parameters.

Two Products, One Mission

The platform's two products serve different needs:

• Web platform: 40+ cloud-hosted frontier models, Study Mode with page-locked answers, Agent Mode with multi-step autonomous task execution, Quiz Hub, unlimited PDF/image uploads, and 5M tokens per active day.
• LumiChats Offline (Windows): CPU-only local inference via GGUF models, LocalDocs (RAG over local files), zero telemetry, no cloud dependency, and bundled fine-tuned models (LumiChats v1.1 3B, LumiChats Coder v2.1 2B, LumiChats 4Bz v1.4 4B, LumiChats TTS 1B, plus two more on HuggingFace).

Study Mode and the Student-First Design

Study Mode is the platform's most distinctive feature. Users upload a PDF and ask questions scoped to specific page ranges; the AI is constrained to answer only from those pages and cites the exact page number in every response. Quiz generation is similarly page-locked. The founder's About page describes the origin as a personal frustration with paying for 30-day AI subscriptions during 8–12-day exam and hackathon bursts.

Open-Source Models and LumiChats Labs

Beyond the platform, LumiChats Labs publishes fine-tuned open-source models on Hugging Face across conversational AI, function calling, medical vision (radiology), OCR/LaTeX extraction, text-to-speech, and instruction following. The homepage lists models including LumiChat Llama (fine-tuned Llama 3.2), LumiFn Qwen (Qwen2.5 for tool use), LumiRadiology, LumiOCR, LumiVoice, and LumiInstruct (Phi-4). The company self-reports download counts in the thousands for individual models.

Update: LumiChats Offline v1.1 and May 2026 Releases

The offline desktop app shipped at v1.1 as the latest release listed on the docs page. According to the About page timeline, May 2026 saw Deep Research Mode, scanned PDF analysis, and an updated Privacy Policy ship. LumiChats Offline reached #7 on Product Hunt's global day leaderboard in May 2026 with 127 upvotes. The web platform added Claude Sonnet 5 (Anthropic's latest) as noted in the homepage banner, signaling active model roster updates.

Who Built It

LumiChats was founded by Aditya Kumar Jha, an engineering student and hackathon participant, who built the platform from scratch — backend, AI pipeline, infrastructure, and growth. Shikhar Burman serves as Co-Founder and CTO, responsible for platform architecture. The company describes itself as independent, founder-owned, and profitable since early months with no outside funding. The self-assessed valuation as of June 2026 is approximately $1.2M–$1.5M, which the founders explicitly note is a founder self-assessment, not an audited figure.