> ## Content Index
> Fetch the complete content index at: https://ping2prompt.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Confluent adds IBM time-series models as built-in Flink functions
- URL: https://ping2prompt.com/confluent-ibm-time-series-models-flink-functions/
- Published: 2026-11-18T15:37:10.000Z
- Updated: 2026-11-18T17:07:37.000Z
- Description: IBM's time-series foundation models now run natively in Confluent as built-in Flink SQL functions, in the cloud and on premises. Teams can forecast and detect anomalies in streaming data on CPUs. Test the functions on one of your own series first.
- Author: Ping2Prompt Newsroom
- Tags: Hugging Face, AI, #short-ping

## IBM time-series models now built into Confluent

IBM's time-series foundation models now run natively inside Confluent as built-in functions, both in the cloud and on premises. The announcement says the models have more than 50 million downloads.

## Forecasting and anomaly detection from Flink SQL

With Apache Flink, teams can apply forecasting, anomaly detection and semantic intelligence to streaming data using SQL. The post describes two functions:

- **`AI_FORECAST`** uses a pretrained model to forecast demand across products and stores; the example gives a 28-day forecast with a P10–P90 range, refreshed as new sales data arrives.
- **`AI_DETECT_ANOMALIES`** builds an expected range for each series from its recent history and compares each new reading against it at a confidence level the team chooses.

Anomaly detection needs no separate model per series. IBM says a pretrained model gives a common starting point across many series, and teams can fine-tune it or combine its output with other data and business logic. The family has four models; the post names the TTM model in its manufacturing example.

## The use cases are scenarios, not measured results

The post's examples are manufacturing quality control and setpoint optimization, retail sales anomaly detection, store-level demand forecasting and reorder points, and card-payment fraud signals. These are illustrative scenarios, not reported results. In the manufacturing case, TTM reads SCADA telemetry and QA measurements every 10 minutes and forecasts the share of tubs in spec four hours ahead.

## Running the models on CPUs in Cloud or on premises

In Confluent Cloud the models are fully managed, with nothing to deploy. On premises they run in your own infrastructure through Confluent Platform for Apache Flink. The models run on CPUs without requiring GPUs. IBM says they are small enough to run on standard CPUs next to the streaming pipeline, which can reduce the need for dedicated GPU infrastructure.

## Who should try in-stream forecasting

If you already run Flink SQL on Confluent and keep sensor, sales or payment series, these functions let you test forecasting without standing up a separate model service. Try them on one of your own series and check the output against known history before you rely on it.

## What IBM and Confluent have not stated yet

The post as provided does not state pricing, licensing, availability dates or which Confluent releases are needed, and the text is cut off. Check the Confluent and IBM documentation before planning a rollout.

## Sources

- [huggingface.co/blog/ibm-research/time-series-powered-streamhouse](https://huggingface.co/blog/ibm-research/time-series-powered-streamhouse?ref=ping2prompt.com)

---

This story was produced with AI assistance from the sources listed above and fact-checked against them before publication. Read our [editorial methodology](https://ping2prompt.com/about-how-we-work/) and the [corrections log](https://ping2prompt.com/corrections/). Spotted an error? Email [editor@ping2prompt.com](mailto:editor@ping2prompt.com).