At Nixtla, we're committed to pushing the boundaries of time series forecasting and making state-of-the-art algorithms accessible to everyone. With TimeGPT-1, we proved for the first time that pre-trained foundation models were possible for time series.
Today we're announcing the private preview of TimeGPT-2 Mini, TimeGPT-2, and TimeGPT-2 Pro. These are the latest versions of our enterprise-grade models, built to reliably solve mission-critical time-series problems. The TimeGPT-2 family is optimized for enterprise needs, prioritizing accuracy and stability with a privacy-first approach and full support for self-hosted and on-premises deployments.
Modular and Purpose-Built
TimeGPT 2 is not a single model, but a modular family designed for real world forecasting.
Each variant is built on a shared foundation and fine tuned for different accuracy latency tradeoffs and hardware environments, enabling flexible deployment across industries and scales.
We are introducing three models, each designed to meet specific operational requirements:
timegpt-2-mini– ideal for fast inference on resource constrained environments.timegpt-2– the best balance between compute cost and accuracy.timegpt-2-pro– the flagship model offering maximum accuracy across short and long horizons and multiple frequencies.
This modular approach allows organizations to choose the model that best fits their operational constraints, without sacrificing accuracy or ease of deployment.
Proven Performance
TimeGPT-2 marks a new milestone in time-series modeling. After extensive testing across diverse enterprise use cases, the TimeGPT-2 family demonstrates up to 60% accuracy improvement compared to the previous generation.
We've also conducted exhaustive benchmarking against public baselines, with TimeGPT-2 consistently ranking in the top 3 on industry-standard benchmarks including:
- GiftEval – evaluating forecasting quality across multiple domains
- FEV – measuring forecast error variance
- VN1 – assessing prediction accuracy
Reproducible results are available upon request. This level of performance is already delivering real value for Fortune 1000 companies across retail, logistics, finance, energy, and IoT sectors.
Enterprise-Ready for Production
TimeGPT-2 is designed to deploy seamlessly into production.
- Deploy on-premise in minutes with a single line of code
- Scale effortlessly to millions of time series with multi-GPU and CPU inference
- SOC 2 ready, meeting the security and compliance needs of enterprises
This combination of accuracy, flexibility, and operational readiness makes TimeGPT-2 a powerful foundation for forecasting in retail, logistics, finance, energy, IoT, and beyond.
Get Early Access
We're currently launching pilot programs with selected organizations. If you'd like to be among the first to leverage TimeGPT-2 in production, join our waitlist to request early access.
This announcement is the first of three releases rolling out in the coming weeks. Stay tuned for more updates.