hjLabs AI Playground
Starting the on-device runtime…
Gemma 3 270M Instruct is a 270M-parameter ONNX model that generates text a token at a time in response to a prompt, streaming its reply as it decodes. On this page it runs entirely inside your browser: the 348 MB download comes straight from the Hugging Face repository “onnx-community/gemma-3-270m-it-ONNX” into your browser cache, and every inference after that executes on your own GPU through WebGPU. Expect a download of roughly a minute on a decent connection. It fits comfortably in the browser cache and in the memory of any GPU from the last few years, including laptop integrated graphics. The model is multilingual, so it is not restricted to English input.
Google’s smallest instruction-tuned Gemma; tight, obedient answers, and a 20 MB tokenizer you only fetch once. In practice people reach for a model like this for drafting, rewriting, summarising a pasted passage, or simply having a conversation that no server ever sees.
Available precisions: q4 (4-bit) and q8 (8-bit). q4 comes first because ONNX Runtime executes 4-bit block-quantised matmuls natively on WebGPU, which is what makes decoding fast enough to be usable — it is a speed choice at least as much as a size one. Hugging Face records about 1k downloads a month. It is a specialist choice rather than a default, which is exactly why it is worth trying when the popular model gets your particular input wrong.
Because there is no inference server behind this page, there is nothing to rate-limit and nothing to bill: the only cost of running Gemma 3 270M Instruct here is the one-off 348 MB download, and once that is cached the page keeps working with the network switched off.
gemma 3 270m instructgemma 3 270m instruct onnxgemma 3 270m instruct browserrun gemma 3 270m instruct locallyonnx-community/gemma-3-270m-it-onnxchat modelgemma 3 270m instruct webgpugemma 3 270m instruct download sizefree chat modeltransformers.js model
About 348 MB for the q4 build, including the tokenizer and config files the runtime also has to fetch — this is the whole payload, not just the weights file. It is downloaded once and then served from the browser Cache Storage API.
Yes, through the WebAssembly fallback, which is slower but works anywhere. WebGPU (Chrome or Edge 113+, Safari 26+) is what makes it feel interactive. On WASM, prefer the smaller quantized build.
No. onnx-community/gemma-3-270m-it-ONNX is a public repository, so your browser fetches the ONNX files directly over HTTPS. There is no account, no key and no server in between.
No. The model weights are downloaded from Hugging Face to your browser once, and every inference after that runs on your own GPU through WebGPU. Your text, images and audio are never sent anywhere — hjLabs.in has no inference server and no way to see your input.