Elastic launches GPU-accelerated inference service for AI workflows

Elastic (NYSE: ESTC) announced the launch of Elastic Inference Service, a GPU-accelerated inference platform integrated into its Elastic Cloud offering. The service targets semantic search, vector search, and generative AI applications.

The new service uses NVIDIA GPUs to provide inference capabilities directly within Elasticsearch. Elastic's built-in sparse vector model, Elastic Learned Sparse EncodeR (ELSER), is available in technical preview as the first text-embedding model on the platform.

"Inference at scale is incredibly important for vector search, semantic search and GenAI workflows," said Steve Kearns, General Manager, Search at Elastic. "The Elastic Inference Service meets that challenge by providing our customers with an API-based inference service using NVIDIA GPUs with our best-in-class Elasticsearch vector database for low-latency, high-throughput inference."

The service operates on consumption-based pricing, charging per model per million tokens. Elastic states the GPU-accelerated system provides up to 10x higher throughput for data ingestion compared to CPU-based alternatives.

Additional models for multilingual embeddings and reranking will be added to the service. The company plans to include models from its recent Jina acquisition in future updates.

The Elastic Inference Service is currently available on Serverless and Elastic Cloud Hosted deployments across all cloud service providers and regions. The service maintains backward compatibility with existing Elasticsearch ML Nodes and supports integration with third-party services through the Open Inference API.

Elastic provides intellectual property indemnity for all models offered through the service, according to the company's announcement.

You May Also Be Interested In

• Scribe Therapeutics files for proposed Nasdaq IPO
• André Branch resigns from Signet Jewelers board
• Seer receives buyout proposal from its own CEO at $2.45 per share