hjLabs AI Playground
Starting the on-device runtime…
mDeBERTa v3 XNLI is a 279M-parameter ONNX model that scores text against categories you invent at query time by reframing the problem as natural-language inference. On this page it runs entirely inside your browser: the 359 MB download comes straight from the Hugging Face repository “Xenova/mDeBERTa-v3-base-xnli-multilingual-nli-2mil7” 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.
Zero-shot labels in 100 languages — write the labels in one language and classify text in another. In practice people reach for a model like this for intent routing, topic labelling and triage where there is no labelled training set and never will be.
Available precisions: q8 (8-bit) and fp32 (full precision). q8 comes first because it roughly halves the download at a quality cost you will struggle to measure on this class of model; fp32 is there when you want the reference numbers. Hugging Face records about 3k 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 mDeBERTa v3 XNLI here is the one-off 359 MB download, and once that is cached the page keeps working with the network switched off.
mdeberta v3 xnlimdeberta v3 xnli onnxmdeberta v3 xnli browserrun mdeberta v3 xnli locallyxenova/mdeberta-v3-base-xnli-multilingual-nli-2mil7zero-shot text labels modelmdeberta v3 xnli webgpumdeberta v3 xnli download sizefree zero-shot text labels modeltransformers.js model
About 359 MB for the q8 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. Xenova/mDeBERTa-v3-base-xnli-multilingual-nli-2mil7 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.