langcache-embed-v3-small, Mellum2-12B-A2.5B-Thinking, and LightOnOCR-2-1B models now available on Amazon SageMaker JumpStart

Amazon SageMaker JumpStart now hosts three new foundation models—Redis’s langcache-embed-v3-small, JetBrains’ Mellum2-12B-A2.5B-Thinking, and LightOn’s LightOnOCR-2-1B—expanding the range of AI tools available to AWS customers. These models address distinct use cases: semantic caching, code reasoning, and document OCR, respectively, enabling more specialized and efficient AI deployments on AWS infrastructure.

langcache-embed-v3-small is designed to optimize semantic caching in large language model applications by converting text into dense vectors that identify equivalent queries. This reduces redundant LLM calls, improving response times in high-volume inference workloads without altering the underlying model architecture or requiring complex pipeline adjustments.

Mellum2-12B-A2.5B-Thinking, developed by JetBrains, targets code generation, debugging, and multi-step reasoning tasks using a Mixture-of-Experts architecture. It activates only 2.5 billion of its 12 billion parameters per inference, supporting a 131,072-token context length and emitting explicit reasoning traces before delivering final answers.

LightOnOCR-2-1B provides end-to-end multilingual document-to-text conversion for PDFs, scans, and images, eliminating the need for traditional OCR pipelines. The 1-billion-parameter model directly converts page images into structured text, achieving high accuracy on OlmOCR-Bench while maintaining a smaller footprint and faster processing than comparable solutions.