{"slug": "aws-launches-sagemaker-training-sessions-for-custom-ai-model-training", "title": "AWS launches SageMaker Training Sessions for custom AI model training", "summary": "Amazon Web Services announced general availability of Amazon SageMaker AI Training Sessions on October 22, 2026, a serverless feature that lets developers run their own post-training loops — including supervised fine-tuning, reinforcement learning, custom loss functions and multi-turn RL — while AWS manages GPUs, distributed execution and infrastructure scaling. Access runs through Python SDK clients from a local machine or SageMaker Studio, with direct calls to forward_backward and optim_step, and the feature supports Llama, Gemma, Qwen, Nemotron and gpt-oss in US East (N. Virginia), US West (Oregon) and Asia Pacific (Mumbai). Pricing is usage-based and tied to tokens processed, and the launch follows serverless full fine-tuning in August 2026 and custom sequence length support on October 21, 2026.", "body_md": "Amazon official logo (public domain, Wikimedia Commons) — CryptoBriefing brand treatment\n\n# AWS launches SageMaker Training Sessions for custom AI model training\n\nAmazon's new serverless feature lets developers write their own training loops while AWS handles the GPUs\n\nAmazon Web Services wants developers to stop babysitting GPUs. On October 22, 2026, the company announced general availability of [Amazon](https://cryptobriefing.com/markets/amazon/) SageMaker AI Training Sessions, a serverless feature for customizing AI models.\n\nThe pitch is simple. Developers write the training logic, and AWS manages the hardware, the scaling and the distributed plumbing underneath.\n\n## What Training Sessions actually does\n\nIn technical terms, Training Sessions lets developers run their own post-training loops. Supported approaches include supervised fine-tuning (SFT), reinforcement learning (RL), custom loss functions and multi-turn RL.\n\nFor the uninitiated, supervised fine-tuning means showing a model examples of correct answers. Reinforcement learning means rewarding it when it gets things right, so it learns which behaviors pay off. Multi-turn RL extends that reward process across longer, back-and-forth interactions, which is closer to how agents operate in practice.\n\nMeanwhile, AWS handles GPU management, distributed execution and infrastructure scaling. The developer keeps algorithmic control without inheriting the operations headache.\n\nAccess runs through Python SDK clients, either from a local machine or from SageMaker Studio. Developers can call session operations directly, including **forward_backward** and **optim_step**.\n\nThose two names describe the core rhythm of model training. One pass computes how wrong the model was and which direction to adjust. The other actually applies the adjustment. Exposing them directly gives developers fine-grained control over each step of the loop.\n\n### AI, tech, and the markets they move—in one daily briefing.\n\nDaily. Free. Join 34,000+ readers across crypto, finance, and policy.\n\nAWS is also shipping a companion “SageMaker training sessions cookbook” with ready-made recipes for SFT and RL, aimed at developers across experience levels.\n\n## Models, regions and the bill\n\nThe feature supports a range of popular open-weight models. The list includes Llama, Gemma, Qwen, Nemotron and gpt-oss.\n\nAt launch, Training Sessions is available in US East (N. Virginia), US West (Oregon) and Asia Pacific (Mumbai).\n\nSecurity features include automatic encryption at rest in S3, with server-side encryption that can use customer-managed KMS keys. Access is governed through IAM managed policies.\n\nPricing is usage-based and tied to the volume of tokens processed, so customers pay roughly in proportion to how much training work they actually run.\n\n## How it fits into AWS’s serverless push\n\nTraining Sessions is not arriving in isolation. It builds on earlier serverless customization features inside SageMaker.\n\nServerless full fine-tuning arrived in August 2026. Support for custom sequence lengths followed on October 21, 2026, just one day before Training Sessions went generally available.\n\nFull fine-tuning covered teams that wanted a more packaged experience. Training Sessions targets the more hands-on crowd, the ones who want to write custom loss functions and control every optimizer step without owning a cluster.\n\n**Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/aws-launches-sagemaker-training-sessions-for-custom-ai-model-training", "canonical_source": "https://cryptobriefing.com/aws-sagemaker-training-sessions-custom-ai/", "published_at": "2026-10-22 21:14:34+00:00", "updated_at": "2026-10-22 21:40:01.877727+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "ai-infrastructure", "ai-tools", "large-language-models"], "entities": ["Amazon Web Services", "Amazon", "SageMaker AI Training Sessions", "SageMaker Studio", "Llama", "Gemma", "Qwen", "Nemotron"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/aws-launches-sagemaker-training-sessions-for-custom-ai-model-training", "markdown": "https://wpnews.pro/news/aws-launches-sagemaker-training-sessions-for-custom-ai-model-training.md", "text": "https://wpnews.pro/news/aws-launches-sagemaker-training-sessions-for-custom-ai-model-training.txt", "jsonld": "https://wpnews.pro/news/aws-launches-sagemaker-training-sessions-for-custom-ai-model-training.jsonld"}}