Elastic Inference Service
Hosted inference endpoint that runs Elastic-managed LLMs, the ELSER sparse-embedding model and third-party embedding models for ingest, search and chat without provisioning ML nodes in a customer's own Elasticsearch deployment.
Find alternatives to Elastic Inference Service
What it does
- Model inference
- Model hosting
- Vector search
Features 5
- Default inference endpointsDeployments include preconfigured default inference endpoints that are ready to use.
- Custom EIS endpointsCreate custom endpoints to instantiate a specific model version or configuration not covered by the defaults.
- Kibana inference managementKibana provides interfaces for managing EIS models and endpoints.
- Region preferencesSettings for where EIS processes inference requests.
- Self-managed cluster access via Cloud ConnectEIS can be used from self-managed clusters through Cloud Connect.
How it compares
Sources
- Pricing
-
Official documentation · 4 Sep 2026
"All models on EIS incur a charge per million tokens … EIS is billed per million tokens used". A rate card per model is not printed on this page.
- Description
-
Official documentation · 4 Sep 2026
Docs: EIS lets you "use machine learning models for ingest, search, and chat independently of your Elasticsearch infrastructure"; lists Elastic Managed LLMs, ELSER on EIS and jina-embeddings-v3.
- Constraints
-
Derived from the security facts on this page, from the same source.
- Security
-
Platform-wide Elastic trust centre list for Elastic Cloud. Omitted: SOC 2 (type not stated), FedRAMP (scope/offering not stated), HIPAA (listed as a standard, no BAA stated). Training, ZDR, residency regions, features not read from this page.
- Sold within
-
Official documentation · 29 Sep 2026
The docs call it "a global service on Elastic Cloud", metered separately from ML-node VCUs: "All models on EIS incur a charge per million tokens."
- Interfaces
-
Official documentation · 12 Oct 2026
Page: 'Alternatively, you can use inference APIs, as described in the following section'.
- Modalities
-
Official documentation · 1 Oct 2026
Page: 'Both models support multimodal embeddings for text, images, video, audio, and documents such as PDF in one shared vector space'.
- Intended for
-
Docs page for engineers: 'you can use inference APIs'; Kibana admin UI requires 'Inference Endpoints: all' privilege to configure endpoints
Also from Elastic
-
Elastic AI Assistant Platform
A conversational assistant embedded in Kibana that answers natural-language questions against a customer's own indexed data across Elastic's Observability, Security and Search solutions.
-
Elastic Agent Builder Agent platform
A builder for custom AI agents that answer questions and take actions over data indexed in Elasticsearch, using configurable tools, skills and prompts.
-
Elastic AI SOC Engine Application
An AI security-operations layer that correlates alerts from a customer's existing security tools, prioritizes threats and guides response workflows without replacing their SIEM.
-
Elastic Workflows Agent platform
An automation engine that runs both scripted steps and AI agents which reason through investigations and execute response actions against data in Elasticsearch; also correlates logs, metrics and traces automatically.
-
Elastic AIOps Application
GenAI- and ML-driven capability inside Elastic Observability that automatically detects, diagnoses and helps resolve operational issues, providing recommended actions for SREs; also answers natural-language questions grounded in observability data, ML jobs and runbooks.
-
Elastic LLM Observability Application
Monitoring capability inside Elastic Observability for generative-AI and agentic applications: performance, cost control, guardrail tracking and reliability for GenAI workloads.
-
Elastic Security Application
Agentic security-operations platform unifying SIEM, XDR and native automation, with autonomous agents handling detection-to-response workflows and purpose-built AI skills for threat hunting, alert analysis and detection engineering; supports multiple LLMs including on-premises models.
-
Elasticsearch Platform
Stores structured, unstructured, and vector data.
-
Elastic Observability Platform
Uses machine learning to identify anomalies before they become incidents.
-
Streams Application
Uses AI to structure and enrich raw logs.
-
Elastic Context Engine Infrastructure service
Distills source data into context for retrieval by agents and applications; also extracts entities such as accounts, contacts and projects, with their relationships.
-
Attack Discovery Application
An LLM-based application that analyzes security alerts, identifies potential attacks, and groups related alerts into attack narratives.
Something wrong here? Send a correction — quote this product id: elastic-inference-service.