YOUR AI WORKSPACE

What can we create today?

Ask questions, explore ideas and get things done. Your all-in-one AI workspace starts here.

SYNORA AI can make mistakes. Verify important information. Chat history is stored only in this browser.

OUR OWN MODEL · v0.2.0

SYNORA CORE

Our original, from-scratch language model. This small research model runs locally in your browser.

OWN MODEL · TRAINED FROM RANDOM INITIALIZATION

SYNORA CORE v0.2

A compact 2-layer causal Transformer with a UTF-8 byte tokenizer. Model weights are published as static assets and inference runs on this browser; prompts are not sent to Gemini or another model provider by this demo.

Architecture2-layer causal Transformer

TokenizerUTF-8 byte-level

Parameters122,688

Training pipelineQueued

Trained weightsPending

Browser inferenceWaiting for weights

Training steps—

Training loss—

Validation loss—

Try our own model

Experimental output only. The starter dataset is small, so answers may be incomplete, repetitive or incoherent. Do not use for important decisions.

Weights load when you run the first prompt.

Your model output will appear here.

Development files

The GitHub Actions pipeline trains the model from random initialization, evaluates held-out examples, exports browser-ready weights and publishes them with the website assets.

Training runs ↗Open Colab ↗View source ↗

COMMUNITY COMPUTE · PILOT

Volunteer GPU Network

Contribute a small amount of GPU compute only when you explicitly start a volunteer session.

WEBGPU PILOT · SYNTHETIC BENCHMARK ONLY

Your device, your choice.

The first network pilot distributes a fixed 32×32 matrix benchmark. It does not train the AI model, process chat prompts, read personal files or run arbitrary code.

◎

Not participating

No compute is running on this device.

Explicit volunteer consent

Required before requesting a GPU adapter or registering this browser.

Power & device confirmation

I confirm this device is plugged in and cool. Temperature cannot be read by this browser.

Resource limits: This pilot runs only a fixed 32×32 synthetic matrix benchmark (not AI training or chat inference). It runs only while this tab is visible, pauses when hidden, and stops after 5 minutes. It sends only a random worker ID and benchmark result. Browser APIs cannot guarantee temperature or battery safety; keep the device plugged in, stop if it warms up, and disconnect any time.

Network status

CoordinatorChecking…

Browser WebGPUNot checked

Active volunteers—

Verified benchmark tasks—

Chat prompts sharedNever

Background computeDisabled

This is a network connectivity and compute-validation pilot, not distributed model training. The coordinator must be deployed and bound before volunteers can connect.

CREATIVE STUDIO

Image Studio COMING SOON

Create and edit images with AI. This feature will be connected when a sustainable free compute option is ready.

✧

Bring your ideas to life

Image generation, image editing and enhancement are planned for a future release. No paid provider is connected yet.

VIDEO WORKSHOP

Video Studio COMING SOON

Text-to-video and image-to-video are planned for a later release.

▶

Video creation is on the roadmap

Video generation requires substantial GPU capacity. We will activate it only after a reliable compute source is available.

DISCOVERY

AI Search COMING SOON

Research with live web sources and citations will be added after a search provider is configured.

⌕

Search the web with AI

Live web search is not connected yet. SYNORA AI will not pretend to browse the web until a search provider is configured.

YOUR WORK

My Library

Your recent conversations saved in this browser.

▤

Your library is empty

Start a conversation and it will appear here on this device.