# AWS makes three DeepSeek models available through SageMaker JumpStart

_Published Wednesday, November 11, 2026 at 7:14 PM EST · AI · Latest · Tier 2 — Notable_

AWS has added DeepSeek-V4-Pro, DeepSeek-V4-Flash and DeepSeek-V4.1-Flash to SageMaker JumpStart, giving customers a console and Python SDK route to deploy the MIT-licensed, open-weight models.

AWS says all three support a 1M-token context window. V4-Pro activates 49B parameters per token from a 1.6T-parameter model, while V4-Flash activates 13B from 284B. Activating only part of each model limits the parameters used for each token.

V4.1-Flash adds native multimodal support and supports 384K maximum output tokens. AWS says its attention and caching techniques reduce the global cache footprint to roughly 1/4 of V4-Flash's; that comparison concerns cache memory.

## Sources

- [Recent Announcements](https://aws.amazon.com/about-aws/whats-new/2026/01/deepseek-v4-pro-v4-flash-v4.1-flash-on-sagemaker-jumpstart/)

---
Canonical: https://techandbusiness.org/newswire/jjvRi_4BXFpt8aWqc4_WvL
Published: 2026-11-12T00:14:41.497Z
Story chronology: 2026-11-11T23:18:00.000Z
Retrieved: 2026-11-12T01:48:38.836Z
Publisher: Tech & Business (techandbusiness.org)