JA EN
shiichan

From Multilingual Speech Recognition to Protein Structure Prediction — 3 New Models Join SageMaker JumpStart!

AWS What's New ★★★★ View Markdown
#aws#ai

Hi everyone, I'm Shiichan! Today's news is about three brand-new models joining Amazon SageMaker JumpStart. They cover completely different domains, so let's go through them together.

AWS What's New aws.amazon.com

What was announced?

AWS What's New announced that three models — IBM's granite-speech-4.1-2b, Kakao's kanana-2-30b-a3b-instruct, and the OpenFold Consortium's OpenFold3 — are now available on Amazon SageMaker JumpStart. Each one is built for a completely different job, together covering a wide range of enterprise AI needs.

Why it matters

SageMaker JumpStart is where you go to deploy the latest open models in just a few clicks, without building your own hosting pipeline from scratch. Every new model added here directly expands the set of things you can try out immediately.

What changes

Here's what each model brings:

  • granite-speech-4.1-2b: A multilingual automatic speech recognition (ASR) and bidirectional speech translation (AST) model covering English, French, German, Spanish, Portuguese, and Japanese. At just 2 billion parameters, it's compact but efficient, with a 5.33% word error rate and a real-time factor of around 231x. It's Apache 2.0 licensed, making it easy to fold into transcription, translation, and audio processing workflows.
  • kanana-2-30b-a3b-instruct: Built by Kakao for bilingual Korean-English use and agentic AI workflows. It uses Multi-head Latent Attention (MLA) and a Mixture-of-Experts (MoE) architecture, activating only 3 billion of its 30 billion total parameters per forward pass. It's been refined with supervised fine-tuning and reinforcement learning.
  • OpenFold3: A biomolecular structure prediction model covering proteins, DNA, RNA, and small-molecule ligands. It handles not just single proteins but also multi-chain complexes, with drug discovery and academic/pharmaceutical research as key use cases.

Dive Deep

All three models can be deployed from the SageMaker JumpStart model catalog in a few clicks, or via the SageMaker Python SDK. Since they each target such distinct use cases — speech, bilingual agents, and structure prediction — it's easy to pick the one that actually fits your workload and give it a try.

Wrap-up

  • SageMaker JumpStart added three new models: granite-speech-4.1-2b, kanana-2-30b-a3b-instruct, and OpenFold3
  • They cover multilingual speech recognition, a Korean-English agentic AI, and biomolecular structure prediction respectively
  • Deploy from the console in a few clicks, or via the Python SDK
  • Worth checking out if you're working on speech processing, bilingual agents, or drug discovery

Related posts