# Udgam Labs · AI research from first principles

Canonical: https://udgamlabs.com · Updated: 2026-10-28

Udgam Labs is an independent AI research lab. Before we build, we understand why things work, whether we are training a model, fine-tuning one or applying AI to a new problem. Every result we publish comes with honest numbers.

Try it in your browser: [https://udgamlabs.com#try](https://udgamlabs.com#try)

Research index · 10 models

## Small models. Measured progress.

Text to video · 28 Oct 2026

### [Drishya Tiny](https://udgamlabs.com/models/drishya-tiny)

Makes 2-second 256×256 clips from a sentence. A research release: about $21 of GPU time, every failure included.

- 212M generator
- 15 h on one L40S
- CC BY-NC 4.0 weights

Telugu language model · 27 Oct 2026

### [Vemana](https://udgamlabs.com/models/vemana)

A Telugu base model trained live in public. On fresh Telugu text, the best open base model under 1B that we tested.

- 278M parameters
- About 16 h on one H100
- Apache-2.0

Chat model · 22 Oct 2026

### [Laghu](https://udgamlabs.com/models/laghu)

A small English chat model with a thinking switch, trained from scratch for $146 and released with everything open.

- 152M parameters
- $146 pretraining run
- Apache-2.0

Speech to text · 19 Oct 2026

### [Shruti](https://udgamlabs.com/models/shruti)

Writes down Hindi, English and Hinglish, offline, on a phone or in the browser. English words stay in English.

- 29.5 MB
- Runs in the browser
- Apache-2.0

Language model · 17 Oct 2026

### [Gargi](https://udgamlabs.com/models/gargi)

A small English language model that thinks step by step on maths problems and runs comfortably on a laptop.

- 384M parameters
- ≈17B training tokens
- Apache-2.0

Maths word problems · 16 Oct 2026

### [Lilavati](https://udgamlabs.com/models/lilavati)

Solves maths word problems by writing a short Python program and running it.

- 95M parameters
- GSM8K 54.7% first try
- Apache-2.0

Text to image · 12 Oct 2026

### [Chitra Tiny](https://udgamlabs.com/models/chitra-tiny)

Makes 256×256 pictures from an English sentence, on a laptop in about three seconds per image.

- 212M parameters
- 256×256
- Apache-2.0

On-device tool calling · 10 Oct 2026

### [Udgam-384M](https://udgamlabs.com/models/udgam-384m)

Turns requests into function calls, documents into JSON and tasks into shell commands, on the device itself.

- 384M parameters
- Context 4,096 tokens
- Apache-2.0

Research release · 29 Sept 2026

### [Udgam English 125M](https://udgamlabs.com/models/english-125m)

A small English model trained from random weights. It missed its performance target, and we published the full results.

- 125M parameters
- Context 2,048 tokens
- Target missed

Fine-tuned adapter · 22 Sept 2026

### [Udgam Hinglish Commands](https://udgamlabs.github.io/hinglish-commands-0.6b/)

Turns short English and Romanized Hinglish requests into a structured intent and details an app can use.

- LoRA on Qwen3-0.6B
- 54.88% vs 9.90% base
- 8 command domains

Latest research

## Reports and papers.

- Technical report · 28 Oct 2026

    ### [Drishya Tiny: building a small video generator from scratch, failures included](https://udgamlabs.com/research/drishya-tiny)

    Technical report for Drishya Tiny, a 212M-parameter video generator trained from scratch with the frozen Wan2.2 VAE and Flan-T5-Base: the design, the data, five stages with what each measured and what failed, VBench, a controlled test of what moves prompt-following, every dollar, and where it falls short.

    [Release post](https://udgamlabs.com/blog/drishya-tiny)

- Technical report · 27 Oct 2026

    ### [Vemana: a 278M Telugu language model trained from scratch](https://udgamlabs.com/research/vemana)

    Technical report for Vemana, a 278M-parameter Telugu base model trained from scratch: data, tokenizer, model, the training run, a pre-registered test on new Telugu Wikipedia pages against every open base model under 1B we found, where it falls short, and the release.

    [Release post](https://udgamlabs.com/blog/vemana)

- Technical report · 22 Oct 2026

    ### [Laghu: a 152M chat model trained from scratch for $146, released fully open](https://udgamlabs.com/research/laghu)

    Technical report for Laghu, a 152M-parameter English chat model trained from scratch for $146 of pretraining compute: architecture, data, the full training recipe, a pre-registered evaluation against size-class rivals, what did not work, the cost and the limitations.

    [Release post](https://udgamlabs.com/blog/laghu)

- Technical report · 19 Oct 2026

    ### [Shruti: offline speech-to-text for Hindi, English and Hinglish in 29.5 MB](https://udgamlabs.com/research/shruti)

    Technical report for Shruti, a 29.5 MB speech-to-text model for Hindi, English and Hinglish built on Cactus Whistle: the Devanagari vocabulary extension, a full fine-tune with 4-bit quantisation-aware training, openly licensed training data with Hinglish labels, one evaluation harness for every model, measurements on a phone and in the browser, and limitations.

    [Release post](https://udgamlabs.com/blog/shruti)

How we think

## Three principles behind every release.

01

### Understand before we build

We work out why something works (the data, the architecture, the training) before we scale it, whether we are training a model, fine-tuning one or applying AI to a new problem.

02

### Honest numbers

Every result names its benchmark, its settings and what it is compared with, and we publish where we are behind.

03

### Small and efficient by design

We build models that run on a laptop or a phone and cost a fraction of the usual amount to train.

Contact

## Research collaborations and partnerships.

We welcome research collaborations, partnerships and questions about our models. Email is best for anything detailed. Replies usually within 24–48 hours.

---

## 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. Today that means open models such as Gargi, a 384M-parameter language model that comes within 2 points of Qwen2.5-0.5B on three reasoning benchmarks with about 1,000× less training data. Next come products, custom models for businesses, and AI in the physical world. Udgam (उद्गम) is Sanskrit for origin.

Udgam Labs · AI research from first principles · https://udgamlabs.com