sarvamai / translation updated 3 weeks ago

sarvam-translate

Name: sarvam-translate
Author: sarvamai

Sarvam-Translate is an advanced translation model built by Sarvam AI in partnership with AI4Bharat, specifically designed for comprehensive, document-level translation across the 22 official Indian languages, built on Gemma3-4B-IT. It addresses modern translation needs by moving beyond isolated sentences to handle long-context inputs, diverse content types, and various formats. Sarvam-Translate aims to provide high-q...

Params

4.3 B

Context

—

Downloads 30d

60 K

Likes

151

Licence: gpl-3.0 Not gated SAFETENSORS 23 languages View on Hugging Face ↗

Download history

tracking started — chart appears after 7 days of snapshots (4 recorded)

60 K downloads in the last 30 days

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	bf16	8.6 GB	10.6 GB	✅ Runs comfortably

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Specifications

Architecture: Gemma3ForConditionalGeneration
Parameters: 4.3 B
Tensor type: BF16
Licence: gpl-3.0
First seen on the Hub: 2025-06-07
Base model: gemma-3-4b-it
Training datasets: undisclosed
Added to our catalog: 2026-08-03

Family

Base model and the most-downloaded derivatives in the catalog.

Base gemma-3-4b-it google 1.8 M

Compare with

nllb-200-distilled-600M 1.6 M opus-mt-nl-en 1.2 M vntl-llama3-8b-v2-gguf 774 K opus-mt-en-ru 686 K opus-mt-en-de 653 K opus-mt-de-en 455 K

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB