# Amazon SageMaker

> Source: https://aiwiki.ai/wiki/amazon_sagemaker
> Updated: 2026-06-21
> Categories: AI Infrastructure, AI Tools & Products, Machine Learning
> From AI Wiki (https://aiwiki.ai), a free encyclopedia of artificial intelligence. Quote with attribution.

**Amazon SageMaker** is Amazon Web Services' fully managed [machine learning](/wiki/machine_learning) platform for building, training, and deploying models at scale, first launched at AWS [re:Invent](/wiki/re_invent) on November 29, 2017 and rebranded in December 2024 as part of a "next generation" platform that AWS calls "the center for all your data, analytics, and AI" [1][6]. It is the central ML offering in the [AWS](/wiki/aws) ecosystem, integrates with [Amazon S3](/wiki/s3), [Amazon EC2](/wiki/ec2), [IAM](/wiki/iam), [VPC](/wiki/vpc), and [CloudWatch](/wiki/cloudwatch), and has grown from three original capabilities into a family of more than two dozen sub-products spanning data labeling, AutoML, foundation-model training, governance, and a unified analytics-and-AI workspace [3][4][5].

What began in 2017 as a focused trio of capabilities (managed Jupyter notebooks, training jobs, and hosted endpoints) has expanded into more than two dozen sub-products covering data labeling, feature stores, AutoML, no-code interfaces, foundation-model training, model governance, and a unified workspace that combines analytics and AI [3][4]. At AWS re:Invent in December 2024, AWS rebranded the legacy ML platform as **Amazon SageMaker AI** and announced a broader "next generation" of SageMaker that bundles SageMaker AI with new lakehouse, governance, and analytics capabilities under a workspace called Amazon SageMaker Unified Studio [5][6]. Unified Studio reached general availability on March 13, 2025 [30].

## When was Amazon SageMaker launched? (History)

Amazon SageMaker was announced and made generally available on November 29, 2017 at AWS re:Invent in Las Vegas, where then-AWS CEO Andy Jassy presented it as an end-to-end platform for building, training, and deploying machine learning models at scale [1][2]. The launch package included managed Jupyter notebook instances, a small library of built-in algorithms, automatic hyperparameter tuning, native support for popular open-source frameworks, and managed real-time hosting endpoints [1]. AWS pitched SageMaker as a way to remove the heavy lifting of provisioning GPU clusters, installing CUDA drivers, packaging containers, and operating inference servers [3].

Major capabilities have been added through annual re:Invent announcements:

* **2018:** [SageMaker Ground Truth](/wiki/sagemaker_ground_truth) (managed data labeling with active learning) and [SageMaker Neo](/wiki/sagemaker_neo) (compile-once, run-anywhere model compiler) [7][8].
* **2019:** [SageMaker Studio](/wiki/sagemaker_studio) (the first integrated development environment for machine learning), [SageMaker Autopilot](/wiki/sagemaker_autopilot) (transparent AutoML), Experiments, Debugger, and Model Monitor [9][10].
* **2020:** [SageMaker Pipelines](/wiki/sagemaker_pipelines) (CI/CD for ML), [SageMaker Data Wrangler](/wiki/sagemaker_data_wrangler), [SageMaker Feature Store](/wiki/sagemaker_feature_store), [SageMaker Clarify](/wiki/sagemaker_clarify) (bias detection and explainability), and the SageMaker Distributed Data Parallel and Model Parallel libraries [11][12].
* **2021:** [SageMaker Canvas](/wiki/sagemaker_canvas) (no-code ML), Serverless Inference (preview), [JumpStart](/wiki/sagemaker_jumpstart) generally available, Ground Truth Plus, Inference Recommender, and SageMaker Savings Plans [13][14].
* **2022:** SageMaker Model Cards, Role Manager, Model Dashboard, geospatial ML capabilities, and notebook-based shadow tests for inference [15].
* **2023:** [SageMaker HyperPod](/wiki/sagemaker_hyperpod) for foundation-model training, smart sifting, inference cost reductions, and richer JumpStart foundation-model coverage [16][17].
* **2024:** "Next generation" of SageMaker, which renames the legacy product to Amazon SageMaker AI and introduces SageMaker Unified Studio, SageMaker Lakehouse, and SageMaker Data and AI Governance [5][6].
* **2025:** General availability of SageMaker Unified Studio (March 2025) [30]; customization of [Amazon Nova](/wiki/amazon_nova) foundation models in SageMaker AI (July 2025) [32]; custom blueprints for Unified Studio (September 2025) [31]; and at re:Invent 2025 (December), serverless model customization, serverless MLflow, new AI-agent notebooks, and checkpointless and elastic training on HyperPod [33][34][35][37].

## What are the major components of SageMaker?

SageMaker is a family of services sharing security, billing, and metadata layers, not a monolithic product.

### Core

**SageMaker Studio**, launched at re:Invent 2019, is a browser-based IDE that hosts JupyterLab notebooks, a code editor, debuggers, model monitors, and dashboards in a single workspace [9]. The 2023 generation added a refreshed UI, web-based VS Code, and tighter integration with [JumpStart](/wiki/sagemaker_jumpstart) and Pipelines [16].

**Notebook Instances** are persistent EC2 Jupyter hosts that pre-date Studio.

**Training jobs** are short-lived managed clusters that pull data from [S3](/wiki/s3), launch containers across CPU or GPU instances, run a training script, persist artifacts back to S3, and shut down. They support built-in algorithms, prebuilt framework containers, and bring-your-own Docker images. SageMaker handles provisioning, health checks, log streaming to [CloudWatch](/wiki/cloudwatch), spot recovery, and cleanup [3][11].

**Inference (Hosting)** offers four deployment patterns: real-time endpoints for low-latency prediction, asynchronous inference for long-running requests staged in S3, batch transform for scoring large datasets without a persistent endpoint, and serverless inference (preview at re:Invent 2021, GA April 2022), which scales endpoints to zero and bills per millisecond [13][14].

**SageMaker Pipelines**, announced at re:Invent 2020, is a CI/CD workflow orchestrator for [MLOps](/wiki/mlops). It defines a directed graph of processing, training, evaluation, model registration, and deployment steps and produces lineage records linking models to the data and code that produced them [11].

**SageMaker Experiments** tracks runs, parameters, metrics, and artifacts and integrates with Studio and the Pipelines lineage graph. **SageMaker Model Registry** is a versioned catalog of approved model packages that downstream pipelines reference for deployment.

### Data labeling and preparation

**SageMaker Ground Truth**, announced at re:Invent 2018, is a managed labeling service combining human annotators with active-learning automation. Workforces can come from Amazon Mechanical Turk, AWS-vetted vendors, or a customer's private team [7]. It supports text, image, video, and 3D point-cloud labeling.

**SageMaker Ground Truth Plus**, announced at re:Invent 2021, is a turnkey variant in which AWS-managed expert workforces produce labels without the customer designing workflows [13].

**SageMaker Data Wrangler**, announced at re:Invent 2020, is a visual data-preparation tool with more than 300 built-in transformations and connectors to S3, Athena, Redshift, and Snowflake [11]. Recipes export as Pipelines steps, processing-job scripts, or feature-store ingestion jobs.

**SageMaker Feature Store**, also from re:Invent 2020, is a repository for ML features with both an online store (low-latency lookup for inference) and an offline store (S3-backed history for training) [11].

### AutoML

**SageMaker Autopilot**, launched at re:Invent 2019, performs AutoML for tabular classification and regression. Unlike opaque AutoML services, Autopilot generates a leaderboard of candidate models along with the underlying notebooks, so practitioners can inspect feature engineering, model selection, and tuning decisions [10]. It is the engine behind several no-code experiences in [SageMaker Canvas](/wiki/sagemaker_canvas).

### Model monitoring and governance

**SageMaker Model Monitor**, announced in 2019, continuously checks production endpoints for [data drift](/wiki/concept_drift), [concept drift](/wiki/concept_drift), and quality regressions by comparing live traffic against a baseline of the training data [10][12].

**SageMaker Clarify**, announced at re:Invent 2020, computes pre- and post-training bias metrics and produces feature-attribution explanations using SHAP-style techniques. Clarify integrates with Model Monitor to alert on shifts in feature importance and supports [algorithmic fairness](/wiki/algorithmic_fairness) and [model monitoring](/wiki/model_monitoring) work [12].

**SageMaker Model Cards**, introduced at re:Invent 2022, centralize model documentation: intended use, training datasets, evaluation metrics, ethical considerations, and risk ratings. They are auto-populated from training metadata and reviewed by humans before approval [15]. **SageMaker Role Manager** simplifies IAM permission setup for ML personas, and **SageMaker Model Dashboard** gives a single view of every deployed model and its monitoring status [15].

### Specialized

**SageMaker JumpStart** is a model and solution hub bundled with Studio. It offers one-click fine-tuning and deployment of pre-trained models from open-source repositories, AWS partners, and third parties such as Hugging Face, Stability AI, AI21 Labs, Cohere, and Meta, including Llama, Mistral, Falcon, Stable Diffusion, BLOOM, and many task-specific computer-vision and NLP models [17].

**SageMaker Canvas**, announced at re:Invent 2021, provides a point-and-click interface for business analysts to build ML predictions without writing code, backed by Autopilot for tabular models and JumpStart for foundation-model use cases [13][14].

**SageMaker HyperPod**, announced at re:Invent 2023, is described by AWS as "a purpose-built infrastructure for distributed training at scale" for foundation models [16]. Clusters are pre-configured with SageMaker's distributed-training libraries, automatically detect failed accelerators and replace them without restarting the job, and provide checkpointing utilities. AWS says HyperPod "reduces time to train foundation models by up to 40%" by recovering from infrastructure faults without human intervention, with launch adopters including Hugging Face, Perplexity, Salesforce, Stability AI, Thomson Reuters, BMW Group, Booking.com, and Vanguard [16][18]. At re:Invent 2025 AWS added checkpointless training, which recovers from hardware failures by transferring model and optimizer state peer-to-peer between healthy nodes rather than restoring from a saved checkpoint, and elastic training, which expands a running job onto idle accelerators and contracts it when higher-priority work needs capacity. AWS says checkpointless training can sustain more than 95 percent training goodput on clusters with thousands of accelerators [37][38].

**SageMaker Neo**, launched at re:Invent 2018, is a model compiler that takes models trained in TensorFlow, PyTorch, MXNet, ONNX, or XGBoost and compiles them for hardware targets including ARM, Intel, and NVIDIA processors as well as AWS [Inferentia](/wiki/aws_inferentia). AWS reported up to 2x performance improvements with no loss in accuracy at launch [8].

**SageMaker Edge Manager** was a fleet-management service for ML models on IoT devices. AWS announced its end of life on April 26, 2024 and recommended ONNX plus AWS IoT Greengrass V2 [19].

## Built-in algorithms

SageMaker ships with a library of built-in algorithms implemented as managed Docker containers. The catalog includes [20]:

| Algorithm | Task |
|---|---|
| [XGBoost](/wiki/xgboost) | Regression, classification, ranking |
| Linear Learner | Regression and binary or multiclass classification |
| K-Means | Clustering with web-scale mini-batch variant |
| Principal Component Analysis | Dimensionality reduction (randomized and exact) |
| Factorization Machines | Recommendation, click-through prediction |
| k-Nearest Neighbors | Classification and regression |
| BlazingText | Word2vec embeddings, supervised text classification (GPU fastText) |
| DeepAR Forecasting | Time-series forecasting with RNNs |
| Object2Vec | General-purpose embeddings of object pairs |
| IP Insights | Anomaly detection on IP addresses |
| Random Cut Forest | Anomaly detection on streaming time series |
| Latent Dirichlet Allocation | Topic modeling (variational inference) |
| Neural Topic Model | Topic modeling (neural variational inference) |
| Image Classification | ResNet-based image classification |
| Object Detection | SSD with VGG and ResNet backbones |
| Semantic Segmentation | FCN, PSP, and DeepLab v3 backbones |
| Sequence-to-Sequence | Machine translation and summarization |

Most algorithms support both single-instance and distributed training, and many include script-mode entry points so users can customize behavior without writing a container from scratch [20].

## What frameworks and hardware does SageMaker support?

### Frameworks

SageMaker provides AWS-maintained Deep Learning Containers for [TensorFlow](/wiki/tensorflow), [PyTorch](/wiki/pytorch), [MXNet](/wiki/mxnet), Hugging Face Transformers, scikit-learn, XGBoost, and R [3][21]. The Hugging Face containers, from a 2021 partnership, bundle the Transformers, Tokenizers, and Datasets libraries in both training and inference variants. A specialized Hugging Face LLM Inference Container based on Text Generation Inference (TGI) was added in 2023 [21].

The SageMaker Training Compiler applies graph optimizations to PyTorch and TensorFlow models. The SageMaker Distributed Data Parallel (SMDDP) and Distributed Model Parallel (SMP) libraries, both released in 2020, accelerate large-scale training by replacing default communication primitives with versions optimized for AWS networking. AWS researchers reported a 44 percent reduction in BERT pre-training time on 512 GPUs at SC20 using these libraries [22].

### Hardware

SageMaker training jobs and endpoints can run on a wide range of EC2 instance families. CPU options include ml.t3, ml.m5, ml.c5, ml.r5, and Graviton-based families. GPU options include ml.p3 (NVIDIA V100), ml.p4 ([NVIDIA A100](/wiki/nvidia_a100)), ml.p5 ([NVIDIA H100](/wiki/nvidia_h100)), ml.p5e ([NVIDIA H200](/wiki/nvidia_h200)), ml.g4dn (NVIDIA T4), and ml.g5 (NVIDIA A10G). AWS-designed accelerators include ml.inf1 ([AWS Inferentia](/wiki/aws_inferentia), 2019), ml.inf2 (Inferentia2, 2023), ml.trn1 ([AWS Trainium](/wiki/aws_trainium), 2022), and ml.trn2 (Trainium2, GA 2024) [23][24].

The Inf1 family launched alongside Inferentia at re:Invent 2019 and became available in SageMaker hosting in 2020, with the AWS Neuron SDK and SageMaker Neo handling model compilation [23]. Inferentia2 and Trainium followed a similar adoption pattern. AWS announced general availability of Trainium2-based Trn2 instances at re:Invent 2024 along with UltraServer configurations that connect 64 chips with high-bandwidth interconnects [24]. At re:Invent 2025 AWS made [Trainium3](/wiki/aws_trainium)-based Trn3 UltraServers generally available; the company describes Trainium3 as its first 3-nanometer AI chip and says Trn3 UltraServers can connect up to 144 chips and deliver up to 4.4 times the compute, about four times the energy efficiency, and nearly four times the memory bandwidth of Trainium2 UltraServers. The Trn3 family is compatible with SageMaker and SageMaker HyperPod through the AWS Neuron SDK [36].

## How does SageMaker support foundation models and generative AI?

SageMaker plays a dual role in AWS's generative-AI strategy. For managed access to proprietary foundation models, AWS offers [Amazon Bedrock](/wiki/aws_bedrock), a separate service exposing models from Anthropic, AI21 Labs, Cohere, Meta, Mistral, Stability AI, and Amazon's own [Amazon Nova](/wiki/amazon_nova) family through a unified API. For customers who want to host, fine-tune, or fully train open-weights or custom foundation models, SageMaker provides JumpStart, HyperPod, and the underlying training and inference infrastructure [16][17].

Key building blocks include **JumpStart** as a foundation-model hub with hundreds of models including Llama, Mistral, Falcon, Stable Diffusion, and embeddings from [Hugging Face](/wiki/hugging_face) [17]; **HyperPod** for resilient large-scale training with hardware health checks and automatic node replacement [16][18]; **Hugging Face Deep Learning Containers** for fine-tuning and inference with PEFT, LoRA, and Text Generation Inference [21]; and inference optimizations such as model parallelism, KV-cache management, and speculative decoding through SageMaker LMI containers. Inside SageMaker Unified Studio, the Bedrock IDE (formerly Bedrock Studio) sits alongside SQL analytics, notebooks, and ML model development [5][6].

### Amazon Nova customization in SageMaker AI

On July 16, 2025 AWS announced a set of customization capabilities for the [Amazon Nova](/wiki/amazon_nova) foundation models inside SageMaker AI, initially covering Nova Micro, Nova Lite, and Nova Pro across the model training lifecycle [32]. The supported techniques are delivered as ready-to-use SageMaker recipes and include supervised fine-tuning (both parameter-efficient and full fine-tuning), alignment via Direct Preference Optimization and Proximal Policy Optimization, continued pre-training, and knowledge distillation to smaller models [32]. Customized Nova models deploy to [Amazon Bedrock](/wiki/aws_bedrock) for inference with on-demand or provisioned throughput [32]. AWS later extended this so that customized Nova models can also be served through SageMaker's own managed inference: on February 16, 2026 it announced general availability of SageMaker Inference for custom Nova models, allowing Nova Micro, Nova Lite, and the reasoning-capable Nova 2 Lite to be trained with SageMaker Training Jobs or HyperPod and then deployed on SageMaker AI managed endpoints in the US East (N. Virginia) and US West (Oregon) Regions [39].

## How does SageMaker integrate with the rest of AWS?

SageMaker leans on the rest of AWS rather than reimplementing common services. [Amazon S3](/wiki/s3) holds training data, model artifacts, batch-transform inputs and outputs, async-inference payloads, and feature-store offline data. [Amazon EC2](/wiki/ec2) provides the compute substrate for every notebook, training job, and endpoint. AWS Lake Formation, AWS Glue, Amazon EMR, Amazon Athena, and Amazon Redshift handle data preparation upstream of training. [IAM](/wiki/iam) scopes execution roles per job, endpoint, or domain user. [VPC](/wiki/vpc) provides network isolation including no-internet training and PrivateLink endpoints. [CloudWatch](/wiki/cloudwatch) collects logs, metrics, alarms, and dashboards. AWS Lambda triggers inference and event-driven retraining; AWS Step Functions handles visual orchestration when Pipelines is not the right fit; AWS CloudFormation and CDK express infrastructure as code; AWS CodeCommit, CodeBuild, and CodePipeline drive CI/CD; and Amazon EKS and ECS host inference in customer-managed clusters while using SageMaker for training and registry [3].

## How much does SageMaker cost? (Pricing)

SageMaker uses on-demand, per-second billing on the underlying compute, with no minimum charge or upfront commitment. Studio itself is free, but the JupyterServer and KernelGateway apps inside Studio incur per-second charges by instance type. Notebook Instances bill per second while running. Training, Processing, Data Wrangler, and HyperPod jobs bill per second from start to completion, including GPU acceleration. Real-time endpoints bill continuously while running, serverless inference bills per millisecond of compute plus data processed, and asynchronous inference and batch transform bill only during request processing. Feature Store charges for online-store traffic and offline-store storage. Ground Truth bills per labeled object plus workforce cost [25][26].

In April 2021 AWS introduced **SageMaker Savings Plans**, a commitment model offering up to 64 percent discount on eligible SageMaker usage for a one- or three-year commitment, alongside on-demand price reductions of up to 14 percent that took effect April 19, 2021 [26]. Multi-model and multi-container endpoints reduce cost when many small models share infrastructure, and Inference Recommender helps right-size endpoints. The serverless MLflow capability introduced at re:Invent 2025 is offered at no additional charge, billing only for the underlying experiment storage and any tracked compute [34].

## How does SageMaker compare to Vertex AI, Azure ML, and Databricks?

SageMaker competes in the broad market for managed ML platforms. The table below summarizes alternatives.

| Platform | Vendor | Strengths | Trade-offs |
|---|---|---|---|
| Amazon SageMaker AI | AWS | Deep AWS integration, mature MLOps, foundation-model training | Steep learning curve, AWS lock-in |
| [Vertex AI](/wiki/vertex_ai) | Google Cloud | Native TPU support, Gemini integration | Smaller third-party model marketplace |
| [Azure Machine Learning](/wiki/azure_ml) | Microsoft Azure | Tight Azure AD and Fabric integration | Features pivot between Studio variants |
| [Databricks](/wiki/databricks) | Databricks | Unified data and ML on Spark, [MLflow](/wiki/mlflow) heritage | Adds another platform layer |
| Snowflake Cortex | Snowflake | ML and LLM functions inside the data warehouse | Less flexible for custom training |
| [Hugging Face](/wiki/hugging_face) Hub and Inference Endpoints | Hugging Face | Largest open model ecosystem | Less governance and lineage |
| Anyscale, Modal, Replicate | Startups | Lighter than full ML platforms | Not full lifecycle MLOps |
| Self-managed [Kubeflow](/wiki/kubeflow) and MLflow | Open source | Maximum control and portability | Heavy operating burden |

For enterprises already standardized on AWS, SageMaker tends to win on integration breadth and governance maturity. Teams that need cross-cloud portability or want a smaller surface area often prefer the alternatives.

## Who uses SageMaker? (Customers)

AWS case studies cite Thomson Reuters, which built its Enterprise AI Platform on SageMaker to unify multi-account ML environments and migrate more than one hundred legacy models [27]; Intuit, which uses SageMaker and [Amazon Bedrock](/wiki/aws_bedrock) for tax-deduction extraction and generative-AI features in TurboTax and QuickBooks and reports cutting model deployment time from six months to one week [28]; and the National Football League, which built its Next Gen Stats player-tracking platform on SageMaker [29]. AWS named Hugging Face, Perplexity, Salesforce, Stability AI, BMW Group, Booking.com, and Vanguard among HyperPod early adopters [16].

## Criticisms

SageMaker is widely deployed, but it has drawn consistent criticism on several fronts:

* **Complexity.** SageMaker spans dozens of sub-products with overlapping functionality, separate consoles, and inconsistent SDK patterns. The 2024 Unified Studio rebrand, which reached general availability in March 2025, was in part a response to user feedback about cognitive overhead [4][5][30].
* **Cost creep.** Notebook apps, real-time endpoints, and HyperPod clusters bill continuously, which can produce unexpected spend if teams forget to shut idle resources down [25][26].
* **Vendor lock-in.** Pipeline definitions, lineage records, model packages, and Feature Store schemas are SageMaker-specific. Teams that decide to leave AWS often find migration nontrivial.
* **Documentation depth.** Each sub-product has its own documentation tree, so getting an end-to-end mental model requires stitching together blog posts, sample notebooks, and re:Invent talks.
* **Edge Manager retirement.** The April 2024 end-of-life of [SageMaker Edge Manager](/wiki/sagemaker_neo) frustrated customers who had built device-fleet management on the service and forced a migration to AWS IoT Greengrass V2 plus ONNX [19].

## Recent direction

SageMaker's direction in 2024 and 2025 reflects three converging trends. First, foundation-model training is now a first-class workload, supported by HyperPod, P5 and Trn2 instances, and the training compiler [16][24]. Second, AWS is consolidating analytics, data engineering, and AI under one platform: at re:Invent 2024 it announced **Amazon SageMaker Unified Studio** (in preview), **Amazon SageMaker Lakehouse**, and **Amazon SageMaker Data and AI Governance**, while renaming the legacy ML platform **Amazon SageMaker AI** [5][6]. AWS frames the new platform as "the center for all your data, analytics, and AI" [6]. Unified Studio brings tools previously scattered across Amazon EMR, AWS Glue, Amazon Redshift, Amazon Bedrock, and SageMaker Studio into a single workspace and integrates [Amazon Q Developer](/wiki/amazon_q) for code assistance [5][6]. Third, AWS positions SageMaker as the customization layer for [Amazon Nova](/wiki/amazon_nova) and as an open complement to [Amazon Bedrock](/wiki/aws_bedrock) for customers who need full control over weights, fine-tuning data, and inference hardware [5][6].

### 2025-2026 developments

**Unified Studio general availability.** On March 13, 2025 AWS made Amazon SageMaker Unified Studio generally available as a single environment for data and AI development. The GA release brought together Amazon Athena, Amazon EMR, AWS Glue, Amazon Redshift, Amazon Managed Workflows for Apache Airflow, and the existing SageMaker Studio, and added simplified permissions management for bringing existing AWS resources into the workspace [30]. As part of the launch, [Amazon Q Developer](/wiki/amazon_q) became generally available inside Unified Studio for natural-language SQL generation, code suggestions, and ETL assistance, and several [Amazon Bedrock](/wiki/aws_bedrock) capabilities reached general availability there as well, including Bedrock Knowledge Bases, Bedrock Guardrails, Bedrock Agents, and Bedrock Flows [30].

**Custom blueprints.** On September 8, 2025 AWS announced the general availability of custom blueprints for Unified Studio. Custom blueprints are administrator-defined, AWS CloudFormation-backed templates that let organizations configure the infrastructure and IAM policies provisioned for each Unified Studio project, so teams can standardize analytics and ML environments and apply their own managed policies to project roles [31].

**re:Invent 2025.** AWS held re:Invent 2025 in Las Vegas from November 30 to December 4, 2025 and announced several SageMaker capabilities [35]:

* **Serverless model customization** in SageMaker AI, a fully managed fine-tuning experience that does not require the user to provision training infrastructure. It can be driven through the console or SDK, or through a natural-language AI agent (in preview), and supports supervised fine-tuning together with Direct Preference Optimization, Reinforcement Learning from AI Feedback (RLAIF), and Reinforcement Learning from Verifiable Rewards (RLVR). AWS lists Amazon Nova, Meta Llama, Qwen, DeepSeek, and GPT-OSS among the models it can customize [33].
* **Serverless MLflow**, a managed version of the MLflow experiment-tracking integration that scales on demand and removes the need to run and size a tracking server. It works with JumpStart, the Model Registry, and Pipelines, and is offered at no additional charge [34].
* **New SageMaker notebooks with a built-in AI agent.** Announced on November 21, 2025, the new serverless notebook experience in Unified Studio lets users run SQL queries, execute Python, process large-scale data jobs, run ML workloads, and create visualizations in one place. The built-in agent, described by AWS as the SageMaker Data Agent, turns natural-language objectives into multi-step execution plans and generates the required SQL and Python by drawing on AWS Glue Data Catalog metadata and the Amazon DataZone business catalog [37].
* **One-click onboarding** for Unified Studio, which AWS says lets teams begin working with existing datasets in minutes using their current IAM roles instead of a multi-week setup [37].
* **Checkpointless training and elastic training** on [SageMaker HyperPod](/wiki/sagemaker_hyperpod), described above, which respectively recover from failures through peer-to-peer state transfer and rebalance jobs across idle accelerators [37][38].
* **Trainium3 UltraServers** ([Trn3](/wiki/aws_trainium)) became generally available and are compatible with SageMaker and HyperPod; AWS describes Trainium3 as its first 3-nanometer AI chip [36].

Alongside these SageMaker features, AWS introduced the **Amazon Nova 2** model family and a service called **Amazon Nova Forge** at re:Invent 2025. Nova Forge lets customers build customized frontier models from Amazon's early, mid, or post-trained Nova checkpoints blended with their own data; AWS positions it for organizations that need deeper customization than fine-tuning provides [33][35].

## See also

- [Prefix caching (automatic prefix caching)](/wiki/prefix_caching)
- [Self-speculative decoding (LayerSkip)](/wiki/self_speculative_decoding)
* [AWS](/wiki/aws), [Amazon Bedrock](/wiki/aws_bedrock), [Amazon Nova](/wiki/amazon_nova), [Amazon Q](/wiki/amazon_q)
* [AWS Inferentia](/wiki/aws_inferentia), [AWS Trainium](/wiki/aws_trainium)
* [Vertex AI](/wiki/vertex_ai), [Azure Machine Learning](/wiki/azure_ml), [Databricks](/wiki/databricks)
* [Kubeflow](/wiki/kubeflow), [MLflow](/wiki/mlflow), [Hugging Face](/wiki/hugging_face)
* [Machine learning](/wiki/machine_learning), [MLOps](/wiki/mlops)

## References

1. Amazon Web Services, "Introducing Amazon SageMaker," What's New at AWS, November 29, 2017. https://aws.amazon.com/about-aws/whats-new/2017/11/introducing-amazon-sagemaker/
2. R. Miller, "AWS releases SageMaker to make it easier to build and deploy machine learning models," TechCrunch, November 29, 2017. https://techcrunch.com/2017/11/29/aws-releases-sagemaker-to-make-it-easier-to-build-and-deploy-machine-learning-models/
3. Amazon Web Services, "Amazon SageMaker Developer Guide." https://docs.aws.amazon.com/sagemaker/latest/dg/whatis.html
4. Amazon Science, "Amazon SageMaker's fifth birthday: Looking back, looking forward," November 2022. https://www.amazon.science/blog/amazon-sagemakers-fifth-birthday-looking-back-looking-forward
5. AWS press release, "AWS Unveils the Next Generation of Amazon SageMaker, Delivering a Unified Platform for Data, Analytics, and AI," December 3, 2024. https://press.aboutamazon.com/2024/12/aws-unveils-the-next-generation-of-amazon-sagemaker-delivering-a-unified-platform-for-data-analytics-and-ai
6. Amazon Web Services, "Introducing the next generation of Amazon SageMaker: The center for all your data, analytics, and AI," AWS News Blog, December 3, 2024. https://aws.amazon.com/blogs/aws/introducing-the-next-generation-of-amazon-sagemaker-the-center-for-all-your-data-analytics-and-ai/
7. Amazon Web Services, "Introducing Amazon SageMaker Ground Truth," What's New at AWS, November 2018. https://aws.amazon.com/about-aws/whats-new/2018/11/introducing-amazon-sagemaker-groundtruth/
8. Amazon Web Services, "Announcing Amazon SageMaker Neo," AWS re:Invent 2018 announcement. https://aws.amazon.com/blogs/machine-learning/category/artificial-intelligence/amazon-sagemaker-neo/
9. R. Miller, "AWS launches SageMaker Studio, a web-based IDE for machine learning," TechCrunch, December 3, 2019. https://techcrunch.com/2019/12/03/aws-launches-sagemaker-studio-a-web-based-ide-for-machine-learning/
10. J. Simon, "AWS re:Invent 2019 AI/ML recap: Amazon SageMaker," Medium, December 2019. https://medium.com/@julsimon/aws-re-invent-2019-ai-ml-recap-part-2-amazon-sagemaker-fee83f05a1bc
11. AWS press release, "AWS Announces Nine New Amazon SageMaker Capabilities," December 8, 2020. https://www.businesswire.com/news/home/20201208005335/en/AWS-Announces-Nine-New-Amazon-SageMaker-Capabilities
12. Amazon Web Services, "New: Amazon SageMaker Clarify Detects Bias and Increases the Transparency of Machine Learning Models," AWS News Blog, December 8, 2020. https://aws.amazon.com/blogs/aws/new-amazon-sagemaker-clarify-detects-bias-and-increases-the-transparency-of-machine-learning-models/
13. AWS press release, "AWS Announces Six New Amazon SageMaker Capabilities," December 1, 2021. https://press.aboutamazon.com/2021/12/aws-announces-six-new-amazon-sagemaker-capabilities
14. Amazon Web Services, "Amazon SageMaker Serverless Inference: Machine Learning Inference without Worrying about Servers," AWS News Blog, December 2021 (preview) and April 2022 (GA). https://aws.amazon.com/blogs/aws/amazon-sagemaker-serverless-inference-machine-learning-inference-without-worrying-about-servers/
15. Amazon Web Services, "Introducing new ML governance tools for Amazon SageMaker," What's New at AWS, November 2022. https://aws.amazon.com/about-aws/whats-new/2022/11/new-ml-governance-tools-amazon-sagemaker/
16. AWS press release, "AWS Announces Five New Amazon SageMaker Capabilities for Scaling with Models," November 29, 2023. https://press.aboutamazon.com/2023/11/aws-announces-five-new-amazon-sagemaker-capabilities-for-scaling-with-models
17. Amazon Web Services, "Amazon SageMaker JumpStart Foundation Models documentation." https://docs.aws.amazon.com/sagemaker/latest/dg/jumpstart-foundation-models.html
18. F. Lardinois, "Amazon SageMaker HyperPod makes it easier to train and fine-tune LLMs," TechCrunch, November 29, 2023. https://techcrunch.com/2023/11/29/amazon-sagemaker-hyperpod-makes-it-easier-to-train-and-fine-tune-llms/
19. Amazon Web Services, "SageMaker Edge Manager end of life," Amazon SageMaker Developer Guide, April 2024. https://docs.aws.amazon.com/sagemaker/latest/dg/edge-eol.html
20. Amazon Web Services, "Built-in algorithms and pretrained models in Amazon SageMaker." https://docs.aws.amazon.com/sagemaker/latest/dg/algos.html
21. Hugging Face, "The Partnership: Amazon SageMaker and Hugging Face," Hugging Face Blog. https://huggingface.co/blog/the-partnership-amazon-sagemaker-and-hugging-face
22. Amazon Web Services, "New: Data Parallelism Library in Amazon SageMaker Simplifies Training on Large Datasets," AWS News Blog, December 2020. https://aws.amazon.com/blogs/aws/managed-data-parallelism-in-amazon-sagemaker-simplifies-training-on-large-datasets/
23. Amazon Web Services, "Announcing availability of Inf1 instances in Amazon SageMaker for high performance and cost-effective machine learning inference," AWS Machine Learning Blog, April 2020. https://aws.amazon.com/blogs/machine-learning/announcing-availability-of-inf1-instances-in-amazon-sagemaker-for-high-performance-and-cost-effective-machine-learning-inference/
24. Amazon Web Services, "Amazon EC2 Trn2 Instances for generative AI compute." https://aws.amazon.com/ec2/instance-types/trn2/
25. Amazon Web Services, "Amazon SageMaker AI Pricing." https://aws.amazon.com/sagemaker/ai/pricing/
26. Amazon Web Services, "Amazon SageMaker announces a price reduction by up to 14% on instances and SageMaker Savings Plans with up to 64% cost saving," What's New at AWS, April 20, 2021. https://aws.amazon.com/about-aws/whats-new/2021/04/amazon-sagemaker-announces-a-price-reduction-in-instances-and-sagemaker-savings-plan/
27. Amazon Web Services, "Streamline and Standardize the Complete ML Lifecycle Using Amazon SageMaker with Thomson Reuters," AWS Customer Case Studies. https://aws.amazon.com/solutions/case-studies/thomson-reuters-sagemaker-case-study/
28. Amazon Web Services, "Intuit on AWS: Case Studies, Videos, Innovator Stories." https://aws.amazon.com/solutions/case-studies/innovators/intuit/
29. Amazon Web Services, "NFL Next Gen Stats: Sports Cloud Solutions." https://aws.amazon.com/sports/nfl/
30. Amazon Web Services, "Collaborate and build faster with Amazon SageMaker Unified Studio, now generally available," AWS News Blog, March 13, 2025. https://aws.amazon.com/blogs/aws/collaborate-and-build-faster-with-amazon-sagemaker-unified-studio-now-generally-available/
31. Amazon Web Services, "Amazon SageMaker Unified Studio announces the general availability of custom blueprints," What's New at AWS, September 8, 2025. https://aws.amazon.com/about-aws/whats-new/2025/09/amazon-sagemaker-unified-studio-custom-blueprints/
32. Amazon Web Services, "Announcing Amazon Nova customization in Amazon SageMaker AI," AWS News Blog, July 16, 2025. https://aws.amazon.com/blogs/aws/announcing-amazon-nova-customization-in-amazon-sagemaker-ai/
33. Amazon Web Services, "Transform AI development with new Amazon SageMaker AI model customization and large-scale training capabilities," AWS Machine Learning Blog, 2026. https://aws.amazon.com/blogs/machine-learning/transform-ai-development-with-new-amazon-sagemaker-ai-model-customization-and-large-scale-training-capabilities/
34. Amazon Web Services, "Accelerate AI development using Amazon SageMaker AI with serverless MLflow," AWS News Blog, December 2, 2025. https://aws.amazon.com/blogs/aws/accelerate-ai-development-using-amazon-sagemaker-ai-with-serverless-mlflow/
35. Amazon Web Services, "Top announcements of AWS re:Invent 2025," AWS News Blog, December 5, 2025. https://aws.amazon.com/blogs/aws/top-announcements-of-aws-reinvent-2025/
36. Amazon Web Services, "Announcing Amazon EC2 Trn3 UltraServers for faster, lower-cost generative AI training," What's New at AWS, December 2025. https://aws.amazon.com/about-aws/whats-new/2025/12/amazon-ec2-trn3-ultraservers/
37. Amazon Web Services, "New one-click onboarding and notebooks with a built-in AI agent in Amazon SageMaker Unified Studio," AWS News Blog, November 21, 2025. https://aws.amazon.com/blogs/aws/new-one-click-onboarding-and-notebooks-with-ai-agent-in-amazon-sagemaker-unified-studio/
38. Amazon Web Services, "Amazon SageMaker HyperPod features." https://aws.amazon.com/sagemaker/ai/hyperpod/features/
39. Amazon Web Services, "Announcing Amazon SageMaker Inference for custom Amazon Nova models," AWS News Blog, February 16, 2026. https://aws.amazon.com/blogs/aws/announcing-amazon-sagemaker-inference-for-custom-amazon-nova-models/