Products
CrowdGPT - community-trained open LLM using volunteers' GPUs instead of datacenters
concept
CrowdGPT is a proposed open-source, free LLM trained in a distributed way: volunteers contribute their own GPU compute instead of relying on datacenters, receiving ownership stakes in the resulting model. The author describes it as rebranded federated learning with a lightweight central server coordinating contributors. It targets people wanting transparent training data and a more ecological alternative to corporate LLM training.
Visit project ↗73 engagement1 sourceSat, Oct 3, 2026, 04:39 PM
AI Overview
- Volunteers contribute their own GPUs to train an open LLM, gaining ownership of the resulting model
- Structured as federated learning with a lightweight centralized server coordinating contributors
- All training and data are claimed to be fully open and transparent
- Positioned as more ecological than datacenter-based training
- Status: early project pitch, no demo, repo, or traction shown
Entities
Related
Releases·Tool Release
Ollama releases tool for running open models locally
Ollama launched a tool that simplifies running open-source large language models locally on personal computers. The release garnered significant attention on Hacker News, indicating strong community interest in accessible local AI deployment.
Ollama launched a tool that simplifies running open-source large language models locally on personal computers. The release garnered significant attention on Hacker News, indicating strong community interest in accessible local AI deployment.
61 engagement1 sourcehackernewsSun, Jul 19, 2026, 07:59 AM
Products·launched
free-llm-apis: GitHub guide to free LLM API tiers for prototypes and side projects
A GitHub guide helping developers choose LLM APIs when budget-limited. It catalogs providers like Google AI Studio, Groq, Mistral, and OpenRouter, comparing free tiers, trial credits, usage limits, and privacy considerations. It solves the problem of finding usable free LLM access for prototypes and side projects.
A GitHub guide helping developers choose LLM APIs when budget-limited. It catalogs providers like Google AI Studio, Groq, Mistral, and OpenRouter, comparing free tiers, trial credits, usage limits, and privacy considerations. It solves the problem of finding usable free LLM access for prototypes and side projects.
1 engagement1 sourceredditSat, Oct 3, 2026, 10:43 AM
Products·building
Interactive guide to building a GPT-style LLM from scratch
A website that teaches how to build and train a small GPT-style LLM from scratch, based on the author's personal experience of training a model on an NVIDIA 4060. It aims to help developers understand LLM internals rather than just using existing tools.
A website that teaches how to build and train a small GPT-style LLM from scratch, based on the author's personal experience of training a model on an NVIDIA 4060. It aims to help developers understand LLM internals rather than just using existing tools.
3 engagement2 sourcesredditFri, Jul 24, 2026, 09:10 AM
Social·Community
Developer trains 102M-parameter LLM to test Indonesia's AI foundation model prospects
A developer trained a 102M-parameter language model to explore whether Indonesia can build its own AI foundation model, contrasting with reliance on commercial APIs like ChatGPT and Claude. The experiment highlights feasibility and challenges for domestic AI development.
A developer trained a 102M-parameter language model to explore whether Indonesia can build its own AI foundation model, contrasting with reliance on commercial APIs like ChatGPT and Claude. The experiment highlights feasibility and challenges for domestic AI development.
0 engagement1 sourcerssMon, Aug 17, 2026, 01:29 AM
Developments·Paper
Govern the Model, Not Only the Data: Storage, Circulation, and Learning in Creative AI
An arXiv paper, 'Govern the Model, Not Only the Data,' argues federated learning's privacy framing is insufficient: distributing computation keeps data on-device while the trained model remains with whoever convened training. The authors contend governance of both data and model — and agency over shaping practices — determines whether AI is extractive, not federation itself.
An arXiv paper, 'Govern the Model, Not Only the Data,' argues federated learning's privacy framing is insufficient: distributing computation keeps data on-device while the trained model remains with whoever convened training. The authors contend governance of both data and model — and agency over shaping practices — determines whether AI is extractive, not federation itself.
0 engagement1 sourcearxivThu, Sep 3, 2026, 01:04 PM