NVIDIA Nemotron-pretraining-Indic-v1: 2.8T-token open Indic pretraining corpus
NVIDIA with IIT Madras's AI4Bharat released a ~2.8 trillion-token pretraining corpus covering 22 Indian languages plus the open GPU-accelerated pipeline used to generate it.
Why it matters
One of the largest open Indic corpora to date, and the full generation recipe (prompts, LLM workflows, quality checks) ships openly so teams can reproduce and extend it.
What you could build with it
An indie developer or small NLP team could continue-pretrain a compact open model on this corpus to build a genuinely strong voice assistant in underserved languages like Odia or Assamese, where English-centric models still fail.
Does it hold up?
Too early to judge — the corpus is announced as releasing soon, not yet downloadable; NVIDIA reports continued pretraining lifted Nemotron 3.5 Super VL 4.98-9.82 points on five Indic benchmarks, but the full corpus has not been independently evaluated.
Built with NVIDIA Nemotron-pretraining-Indic-v1
- RuntimeWire: NVIDIA and AI4Bharat announce a 2.8T-token corpus for 22 Indian languagesblog · Writeup of the announcement, noting the token total alone does not establish language-level quality or downstream performance.
- Unite.AI: NVIDIA unveils 2.8T-token Indic pretraining corpus and open pipelineblog · Coverage of the corpus release with details on availability and the stated limitations on translation-based text.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71JetBrains Mellum 2.1: 12B MoE coding model trained for agentsJetBrains released Mellum 2.1, a 12B MoE model (2.5B active parameters, 128K context) under Apache 2.0, trained mainly…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.