Yug-1: Birla AI Labs' Open Time Series AI Model, Explained

Most AI news is about chatbots. Yug-1, from Aditya Birla Group's Birla AI Labs, is about something else: numbers that change over time. Sales by week, machine sensor readings, power use, prices.

What launched

Birla AI Labs announced Yug-1 on Monday, 19 October 2026. It is a time series foundation model, or TSFM. According to The Economic Times, it has 271.8 million parameters, and the model is available on GitHub and Hugging Face for developers and researchers to test.

What is a time series model?

A time series is a list of measurements in order of time: daily demand for a product, a factory machine's temperature every minute, monthly electricity use. A foundation model is trained once on a large, varied collection of such patterns, so it can be used on new data without training a fresh model for each problem.

Birla AI Labs says Yug-1 can forecast outcomes, find patterns and flag anomalies across finance, weather, energy and retail data. The Economic Times reports that it uses zero-shot probabilistic forecasting. Zero-shot means it can be pointed at new data without extra training. Probabilistic means it gives a range of likely outcomes, so you can see how uncertain a forecast is, not just one number.

How it scores

Ananya Birla, founder and chairperson of Birla AI Labs, said in a LinkedIn post that Yug-1 beat other first-generation TSFMs on the GiftEval benchmark and holds its own against newer models. The Economic Times describes the chart she shared. It shows Yug-1 at 0.744 on MASE, an error measure where lower is better.

• Google TimesFM 2.0: 0.758
• Amazon Chronos Bolt Base: 0.808
• IBM TTM-R3-PT: 0.724 (better than Yug-1)
• Salesforce Moirai2: 0.728 (better than Yug-1)

So Yug-1 is ahead of two well-known models and behind two others on that one chart. These are the company's own numbers, and we have not re-run them.

Why it matters in India

Companies already use time series forecasting for demand planning, equipment monitoring and finance. A free, open model that works on new data without retraining could make this easier for small teams and students. The Statesman reports that Birla AI Labs already runs AI programmes across Aditya Birla Group, including credit underwriting at Aditya Birla Capital and factory monitoring at Hindalco. Birla AI Labs calls Yug-1 Asia's first TSFM built for real-world use. That is the company's claim, and we could not verify it independently.

How to try it

1. Search for Yug-1 on Hugging Face or GitHub from Birla AI Labs and read the model card first, including the licence.
2. Take a CSV of something you know well, like weekly sales or daily website visits.
3. Hold back the last few weeks, ask the model to forecast them, and compare with what really happened.
4. Check whether the range it gives is wide or narrow, and whether the real numbers fell inside it.

Do not use a forecast from any model as the only basis for a money decision.

Sources

• The Economic Times, 19 October 2026: Birla AI Labs launches Yug-1 to forecast demand and spot patterns in business data
• The Statesman, 19 October 2026: Birla AI Labs launches Made-In-India time series foundation model Yug-1