What matters in AI.

Subscribe

IBM time-series models operate in Confluent on standard CPUs

The models predict values in live data and use no GPUs.

IBM and Confluent put IBM time-series models into Confluent as built-in functions. Confluent makes software that sends business data between systems as it changes. The models predict the next values of data that changes with time, such as sales or sensor data from a plant or a store. IBM tells that they have more than 50 million downloads.

How it works

Teams use the models with SQL, a database query language. They operate in Apache Flink, an engine that works on data while it flows, in the cloud or on premises. One built-in function gives each sales series a usual range from its recent data.

The proof

The models operate on standard CPUs, with no GPUs. The only number given is more than 50 million downloads. IBM gave no results for speed, precision or cost.

What is not known

It is not known how correct the models are on data of a company. Teams can tune a model or put its output together with business rules and human decisions. The results change with the configuration of the pipeline.

Insight: IBM puts its models where its data flows

Confluent is in the IBM group, and IBM models operate in it. In November 2026, Amazon released Mitra-v2, an open model that predicts from tables with no new training. This reads as a move to hold the point where prediction occurs. The model that is easy to start in the data stream can get most. Leaders who select a prediction tool also select a data platform.

Sources

  1. Time-Series Models in the Streamhouse: A New Frontier and Use Caseshuggingface.co
AI MATTER · NEWS · AI MATTER · NEWS ·18 NOV2026

Posted

Tags

More in Models

All Models news