UDGAM LABS · Translator · English → Hindi · 17.3M parameters, 17.7 MB at int8
Paar English → Hindi translates English sentences into Hindi (Devanagari). It has 17.3M parameters: the int8 model file is 17.7 MB (19.6 MB with its vocabulary), and it runs offline on an ordinary CPU, about 36 ms a sentence on one core of a Mac mini, or in a web browser with transformers.js. It was distilled from IndicTrans2: it learned from English sentences that IndicTrans2 dist-200M translated. Against Firefox's offline English-to-Hindi model (17.1 MB, a similar size), it is behind by 2.64 chrF++ on FLORES+ (news-style text) and behind by 1.87 on IN22-Conv (everyday conversation). Against indicMT, a model about 8.6 times larger to download, it is behind by 1.43 on FLORES+ and behind by 2.30 on IN22-Conv. All differences are paired, with 95% intervals (Results below).
It is one of twelve, English to: Hindi, Bengali, Marathi, Telugu, Tamil, Gujarati, Urdu, Kannada, Malayalam, Punjabi, Odia, Assamese.
Try it: in your browser · code and fine-tuning recipe on GitHub
Transformers (the weights as trained):
# pip install transformers sentencepiece torch
from transformers import MarianMTModel, MarianTokenizer
repo = "udgamlabs/Paar-en-hi"
tok = MarianTokenizer.from_pretrained(repo)
model = MarianMTModel.from_pretrained(repo)
batch = tok(["Can you send me the invoice for the last three months?"], return_tensors="pt", padding=True, truncation=True, max_length=128)
out = model.generate(**batch, num_beams=1, max_new_tokens=200)
print(tok.batch_decode(out, skip_special_tokens=True)[0])
# क्या आप मुझे पिछले तीन महीनों से चालान भेज सकते हैं?
CTranslate2 int8 (the small, fast build in ct2-int8/):
# pip install ctranslate2 sentencepiece huggingface_hub
import ctranslate2, sentencepiece as spm
from huggingface_hub import snapshot_download
path = snapshot_download("udgamlabs/Paar-en-hi", allow_patterns=["config.json", "ct2-int8/*"])
sp = spm.SentencePieceProcessor(model_file=f"{path}/ct2-int8/source.spm")
tr = ctranslate2.Translator(f"{path}/ct2-int8", device="cpu", compute_type="int8")
def translate(sentences):
pieces = [p[:127] + ["</s>"] for p in sp.encode(sentences, out_type=str)]
out = tr.translate_batch(pieces, beam_size=1, max_decoding_length=200)
return [sp.decode([t for t in r.hypotheses[0] if t != "</s>"]) for r in out]
print(translate(["Can you send me the invoice for the last three months?"])[0])
JavaScript, in the browser or Node, with transformers.js (the ONNX build in onnx/, int8):
// npm install @huggingface/transformers
import { pipeline } from "@huggingface/transformers";
const translate = await pipeline("translation", "udgamlabs/Paar-en-hi", { dtype: "q8" });
const [out] = await translate("Can you send me the invoice for the last three months?", { max_new_tokens: 200 });
console.log(out.translation_text);
// क्या आप मुझे पिछले तीन महीनों से चालान भेज सकते हैं?
num_beams=1) is what we scored. Beam search was not tested.| English | Paar English → Hindi (from the Transformers quick start) |
|---|---|
| The train leaves at 6:45 tomorrow morning, so please be at the station by 6:15. | ट्रेन कल सुबह 6.45 बजे रवाना होती है, इसलिए कृपया स्टेशन पर 6.15 बजे तक रहें। |
| Can you send me the invoice for the last three months? | क्या आप मुझे पिछले तीन महीनों से चालान भेज सकते हैं? |
| My phone battery dies by the afternoon even though I charge it overnight. | दोपहर तक मेरी फोन की बैटरी की मृत्यु हो जाती है, भले ही मैं इसे रातोंरात चार्ज करता हूं। |
These are unedited outputs, mistakes included, and no native speaker has reviewed them.
chrF++ (higher is better), NFC on both sides, our harness, the same sentences for every system. Ours greedy; indicMT int8 with beam 5 as its card says; Firefox's offline model greedy, as Firefox runs it. Differences (ours minus theirs) are paired: 1,000 bootstrap resamples, 95% interval. Measured 3 to 16 November 2026. Our scores are of the fp32 weights; the int8 build is within 0.07 chrF++ of them, and the comparison with Firefox, re-run on the int8 build, gives the same result.
| Test set (sentences) | Paar (17.3M) | indicMT (153.8M) | IndicTrans2 dist-200M (211.8M) | Firefox's offline model (17.1 MB) |
|---|---|---|---|---|
| FLORES+ devtest (1,012) | 55.49 | 56.92 (-1.43 [-1.89, -0.97]) | 58.60 | 58.13 (-2.64 [-3.07, -2.16]) |
| IN22-Conv (1,503) | 44.92 | 47.22 (-2.30 [-2.83, -1.81]) | 48.79 | 46.79 (-1.87 [-2.36, -1.37]) |
<unk> in
transformers.js), or are dropped by the Transformers decoder. 2.0% of FLORES+ outputs carry one.A Marian encoder-decoder (6 encoder and 2 decoder layers, width 256, one shared 32,000-piece SentencePiece vocabulary), trained from random weights. Its training targets are IndicTrans2 dist-200M's translations (beam 4) of English sentences from IndicCorp v2 (CC0). The sentences were cleaned and deduplicated, about a third were rewritten in a spoken style, and every sentence overlapping FLORES+, NTREX or IN22 was removed before training. Scores are on the test sets, scored once after training.
For translating short English text into Hindi where a large model is too heavy or the text should not leave the device: apps, web pages, offline tools, browser extensions, a first draft for a human translator. It is not meant for legal, medical or official text without a human check. Check names and numbers in the output.
The GitHub repository has a recipe (finetune/) for training it further on your own
sentence pairs with Transformers. If you publish a fine-tuned version, set base_model: udgamlabs/Paar-en-hi in its card so it
links back here.
Apache-2.0. See NOTICE: IndicTrans2 (AI4Bharat, MIT) wrote the training targets, and the English sentences come from
IndicCorp v2 (AI4Bharat, CC0). No third-party weights are included.
@misc{udgamlabs2026paarhi,
title = {Paar: a 17 MB English-to-Hindi translator},
author = {Prasad, Jitesh K},
year = {2026},
howpublished = {Udgam Labs},
url = {https://huggingface.co/udgamlabs/Paar-en-hi}
}
Please also cite IndicTrans2 (Gala et al., 2023, TMLR; arXiv:2305.16307), whose translations it learned from.
About Udgam Labs. Udgam Labs is an independent AI research lab. We work from first principles: before we build, we understand why things work, whether we are training a model, fine-tuning one or applying AI to a new problem. Udgam (उद्गम) is Sanskrit for origin. udgamlabs.com
model.safetensors 69 MBonnx/encoder_model.onnx 58 MBonnx/decoder_model_merged.onnx 44 MBct2-int8/model.bin 18 MBonnx/encoder_model_quantized.onnx 15 MBonnx/decoder_model_merged_quantized.onnx 11 MB