Amazon SageMaker JumpStart: 2026 year-to-date in review

Amazon SageMaker JumpStart, a machine learning (ML) hub within Amazon SageMaker AI, made notable progress in 2026. The platform serves as a curated catalog of AI models, solution templates, and example notebooks, simplifying the transition from model selection to deployment. It enables users to evaluate open source large language models (LLMs), deploy vision models for document understanding, and fine-tune compact models on their own data.

In terms of onboarding new models, SageMaker JumpStart welcomed over 100 new models across 30 distinct providers from January to September 2026. This expansion was driven by the rapid release of open source models and the increasing customer demand for diverse model architectures, sizes, and capabilities. The leading contributors to this growth included Alibaba (Qwen) with 29 models, NVIDIA with 17 models, Mistral with 7 models, Google with 6 models, IBM with 4 models, and Cohere with 4 models.

Additionally, 33 models were onboarded from other providers such as MiniMax, Zhipu AI, DeepSeek, Moonshot AI, Allen AI, Microsoft, Meta, Reflection, Black Forest Labs, JetBrains, OpenBMB, Kakao, OpenFold, and more.

Some of the notable model onboarding highlights include NVIDIA's Cosmos3 world foundation model family (Super, Edge, Nano), which brings video generation, physics simulation, and action planning to physical AI applications, and NVIDIA Nemotron models spanning from efficient edge inference (Nano-12B) to large-scale reasoning (Super-120B). Google's Gemma 4 variants add vision-language capabilities, while IBM's Granite 4.0 h-micro/small/tiny and Granite-speech-4.1-2b models support multilingual automatic speech recognition.