Introducing Synthefy API: State-of-the-Art Time Series Forecasting for Everyone

Imagine being able to ask:

A utilities company: "Forecast my neighborhood's energy demand during a winter freeze in Austin, assuming 50% of homes own electric vehicles, so I can plan ahead."

A retailer: "Forecast customer demand for my products this Black Friday — optimized for new ad campaigns and website traffic — so I can stock inventory confidently."

And then getting state-of-the-art answers in minutes, not months. That is the power of time series foundation models.

Traditionally, solving problems like these has required 5–6 months of effort from specialized engineering teams — building pipelines, tuning models, and managing GPUs.

Until now.

We're excited to announce the launch of Synthefy's Zero Shot Forecasting API — your gateway to the most advanced forecasting technology available today.

For the first time, you can directly access our generative AI models purpose-built for time series — no engineering team, no GPU cluster, no training pipeline required.

Time Series Everywhere

Time series forecasting drives critical decisions in every industry. Revenue and demand planning, supply chain optimization, energy load prediction, and financial risk modeling all rely on accurate forecasts.

For years, most models were univariate. They only look at one signal at a time and ignore the rich context around it. That limitation was not by choice but by necessity. The modeling tools of the past could not handle multiple interacting variables effectively. There was no reason to collect more data when the models could not use it.

Synthefy's SOTA Forecasting Models

Synthefy's state-of-the-art† models change that equation. They are multivariate and metadata-aware, able to learn from multiple variables simultaneously, and condition on external contexts such as weather, events, financial indicators, and macroeconomic trends.

This means the more relevant data you bring, the better your forecasts become. Data enrichment is no longer a nice-to-have. It is essential for getting the maximum value from our models. And it's built into the Synthefy API. Learn More →

Why Not LLMs?

Large Language Models have revolutionized text processing, but time series is a fundamentally different modality. LLMs struggle with continuous numeric data, with diverse metadata formats, and with preserving structure when everything is forced into a single text stream.

In our tests, LLMs missed obvious spikes and failed to capture subtle but important trends.

Synthefy's models are built from the ground up for time series. They are natively continuous and can integrate arbitrary metadata without lossy conversions. They are built on top of our universal metadata encoder — allowing a single model to work across domains.

Simply put, it's the answer that time series needed. Use our models on Synthefy API. Learn More →

The API Experience

We have made advanced forecasting radically simple:

1. Sign Up

2. Get an API Key

Click "API Keys" > "+ Create API Key"

3. Install the SDK

pip install synthefy

4. Start Forecasting

from synthefy import SynthefyAPIClient

with SynthefyAPIClient() as client:
    forecast_dfs = client.forecast_dfs(
        history_dfs=[history_df],
        target_dfs=[target_df],
        target_col='sales',
        timestamp_col='date',
        metadata_cols=['store_id', 'category_id', 'promotion_active'],
        leak_cols=[],
        model='sfm_moe'
    )
    
forecast_df = forecast_dfs[0]
print(forecast_df[['timestamps', 'sales']].head())

More documentation is available on PyPI and via our example notebooks.

Key Features

• Pay per API call so you can scale usage as needed
• Forecasts in seconds instead of days
• No compute required
• Zero training

The Bottom Line

Synthefy's API puts production-ready forecasting in your hands without needing an infrastructure or data science team.

Synthefy can instantly replace your usage of linear regression, ARIMA, or Prophet, within minutes.

The world of time series modeling has moved beyond univariate approaches. With Synthefy, you can finally take advantage of models that match the complexity of your data and turn richer context into better decisions.

Early Access Sign Up →

† State-of-the-Art result on real customer datasets compared to Chronos, TimesFM, ARIMA, SARIMA, and other time series modeling tools.

Originally published on Medium