Run open-source AI models locally, share them across your network, race two models head-to-head, and learn machine learning by training a small model yourself. One app — Windows, Linux and Android. No cloud bill.
Free · Apache-2.0 open source · fully offline — models run on your hardware, nothing leaves your network
Local AI is powerful but trapped behind command lines. Symposium puts installing, running, sharing, comparing and even training models behind buttons.
One-click model download and streaming chat with open-source LLMs running on your own PC — with live tokens-per-second.
Flip one toggle and your PC serves its models to your friends' phones and laptops on the same network. They find it automatically.
Two panes, any two models on the network. Fire one prompt at both, watch them race, and vote on the winner.
Play with temperature, top-p and system prompts in a live edit-and-retest loop. Share personas as small JSON files.
Train a tiny GPT on your own machine and watch the loss curve fall in real time — then chat with your half-trained model. The best way to feel how ML works.
LEARN.md teaches everything from scratch — tokens, streaming, network discovery, loss curves — one chapter per feature you can see running.
Classmate AI teaches you concepts with Astra. Symposium is where you get your hands dirty: if you're learning machine learning, the moment it clicks is when a model you trained yourself says its first (terrible) sentence. Everything runs on your hardware — no API keys required, no cloud bill, nothing to top up. That's the same cost-first philosophy the rest of Visionary Sparks runs on.
Openly built, early and honest: today devices find each other on the same Wi-Fi automatically; across the internet you can already join by IP if the host forwards a port. The one-code version is next.
Unsigned early-access build — Play Store listing is on the roadmap.
Stuck? Email us — we answer.