Why This Matters
This story matters because it showcases a practical, in-browser demonstration of running a local decision-making language model entirely client-side, with no backend or waitlist required. This approach highlights a growing trend toward privacy-preserving, low-latency AI that runs directly on consumer hardware, letting users measure real performance tradeoffs themselves.
Key Takeaways
- The demo lets users compare reading probability distributions versus generating them token-by-token, directly in-browser.
- It requires no backend infrastructure, no signup, and no waitlist—just a GPU and a browser.
- Model options scale to device capability, with MiniCPM 2B as default and a smaller Qwen3 0.6B option for phones.
Can we run something like Jev in your browser?
A live, local experiment
Decision model in your browser.
A local model can either read probabilities for your allowed options without decoding them, or write the same kind of distribution token by token. Pick a size, run both on your own GPU, and measure the difference.
browser only no backend your timings 1.56 GB model
celebrationThere is no waitlist! Just try it out ↓
MiniCPM5 2B is selected by default. On a phone or smaller device, switch to Qwen3 0.6B in the model box if needed.
GoKawiil