--- title: Real-Time Intelligence with IBM Time Series Models on... description: IBM's Granite time series foundation models are now available in Early Access as native inference on Confluent Cloud, callable directly from Apache Flink SQL... canonical: https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Real-Time Intelligence with IBM Time Series Models on Confluent | daily.dev og:description: IBM's Granite time series foundation models are now available in Early Access as native inference on Confluent Cloud, callable directly from Apache Flink SQL... og:url: https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp og:image: https://api.daily.dev/og/posts/KoRfUoERP.png og:image:alt: Real-Time Intelligence with IBM Time Series Models on Confluent og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Real-Time Intelligence with IBM Time Series Models on Confluent **[Hugging Face](https://daily.dev/sources/huggingface)** · 14 min read · 0 upvotes · 0 comments ## Summary IBM's Granite time series foundation models are now available in Early Access as native inference on Confluent Cloud, callable directly from Apache Flink SQL via AI_FORECAST and AI_DETECT_ANOMALIES functions. The portfolio includes four complementary models (PatchTST-FM, FlowState, TTM, TSPulse) covering forecasting, anomaly detection, similarity search, classification and gap-filling, switchable with a single SQL parameter. Confluent supplies streaming state and governance while IBM's pretrained models eliminate per-series bespoke modeling, with use cases spanning retail demand planning, fraud detection, and manufacturing process optimization. Access starts on Confluent Cloud on AWS, with Confluent Platform for on-prem/hybrid environments planned next. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I run IBM Granite time series forecasting models directly from Flink SQL on Confluent Cloud? Use the AI_FORECAST function with a JSON_OBJECT parameter specifying the model name (such as 'ttm') and horizon, applied over a windowed SELECT on a streaming table ordered by event time. Confluent Cloud runs the model natively within Apache Flink, with no separate ML stack, provider credentials, or model-serving infrastructure to manage. Switching models only requires changing the model parameter value. _daily.dev surfaces streaming SQL patterns like this for teams building real-time forecasting pipelines._ ### What are the differences between PatchTST-FM, FlowState, TTM, and TSPulse time series models from IBM? PatchTST-FM reads a series in patches like a language model and returns a full probability distribution, useful for percentile-based planning. FlowState maintains a continuously updated running summary suited to mixed-frequency data. TTM uses lightweight mixing networks instead of attention, letting a million-parameter model cover a hundred thousand series nightly on CPU. TSPulse combines time and frequency views for anomaly detection, classification, and gap-filling. _Developers comparing time series foundation models can track releases like these on daily.dev before choosing one._ ### How can I run real-time fraud anomaly detection on payment streams without labeled training data? IBM's Granite time series models detect fraud by scoring each payment against a per-card behavioral baseline via the AI_DETECT_ANOMALIES call, flagging drift toward suspicious activity before a transaction completes. Because the model transfers pretrained knowledge to new products and corridors, protection starts without labeled fraud cases, and the model is later refined continuously as confirmed cases arrive. _daily.dev helps engineers evaluating anomaly detection approaches keep up with new streaming ML capabilities._ ## Similar posts on daily.dev - [New in Confluent Intelligence and AI Tools: Making Agents Native to the Stream, Expanded Model Support, New Agent Skills, and Copilot](https://daily.dev/posts/new-in-confluent-intelligence-and-ai-tools-making-agents-native-to-the-stream-expanded-model-suppo-rc1fwyggr) · Confluent Blog · 0 upvotes · 0 comments - [IBM releases SOTA Granite Time Series PatchTST-FM-r2 model with commercial-friendly license](https://daily.dev/posts/ibm-releases-sota-granite-time-series-patchtst-fm-r2-model-with-commercial-friendly-license-xbwkizgkr) · Hugging Face · 0 upvotes · 0 comments - [Confluent Makes it Easier to Build and Secure Real-Time AI at Scale](https://daily.dev/posts/confluent-makes-it-easier-to-build-and-secure-real-time-ai-at-scale-cwztvhsas) · SD Times · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#confluent-cloud](https://daily.dev/tags/confluent-cloud), [#apache-flink](https://daily.dev/tags/apache-flink), [#time-series-forecasting](https://daily.dev/tags/time-series-forecasting) [View this post on daily.dev](https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Real-Time Intelligence with IBM Time Series Models on Confluent","url":"https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp"},"datePublished":"2026-09-02T13:52:22.664Z","dateModified":"2026-09-02T13:57:22.496Z","description":"IBM's Granite time series foundation models are now available in Early Access as native inference on Confluent Cloud, callable directly from Apache Flink SQL...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/481d8516ca0ed7c2099c3818c6da9840?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/481d8516ca0ed7c2099c3818c6da9840?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hugging Face","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hugging Face","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/f1f55c67d81a4330acf5b90b26b0c8e1","url":"https://daily.dev/sources/huggingface"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,confluent-cloud,apache-flink,time-series-forecasting","timeRequired":"PT14M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hugging Face","item":"https://daily.dev/sources/huggingface"},{"@type":"ListItem","position":3,"name":"Real-Time Intelligence with IBM Time Series Models on Confluent"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/real-time-intelligence-with-ibm-time-series-models-on-confluent-korfuoerp#faq","mainEntity":[{"@type":"Question","name":"How do I run IBM Granite time series forecasting models directly from Flink SQL on Confluent Cloud?","acceptedAnswer":{"@type":"Answer","text":"Use the AI_FORECAST function with a JSON_OBJECT parameter specifying the model name (such as 'ttm') and horizon, applied over a windowed SELECT on a streaming table ordered by event time. Confluent Cloud runs the model natively within Apache Flink, with no separate ML stack, provider credentials, or model-serving infrastructure to manage. Switching models only requires changing the model parameter value. daily.dev surfaces streaming SQL patterns like this for teams building real-time forecasting pipelines."}},{"@type":"Question","name":"What are the differences between PatchTST-FM, FlowState, TTM, and TSPulse time series models from IBM?","acceptedAnswer":{"@type":"Answer","text":"PatchTST-FM reads a series in patches like a language model and returns a full probability distribution, useful for percentile-based planning. FlowState maintains a continuously updated running summary suited to mixed-frequency data. TTM uses lightweight mixing networks instead of attention, letting a million-parameter model cover a hundred thousand series nightly on CPU. TSPulse combines time and frequency views for anomaly detection, classification, and gap-filling. Developers comparing time series foundation models can track releases like these on daily.dev before choosing one."}},{"@type":"Question","name":"How can I run real-time fraud anomaly detection on payment streams without labeled training data?","acceptedAnswer":{"@type":"Answer","text":"IBM's Granite time series models detect fraud by scoring each payment against a per-card behavioral baseline via the AI_DETECT_ANOMALIES call, flagging drift toward suspicious activity before a transaction completes. Because the model transfers pretrained knowledge to new products and corridors, protection starts without labeled fraud cases, and the model is later refined continuously as confirmed cases arrive. daily.dev helps engineers evaluating anomaly detection approaches keep up with new streaming ML capabilities."}}]} ```